Here we introduce scBench (read the paper), a benchmark of 394 verifiable problems drawn from real scRNA-seq workflows. It spans six sequencing platforms, seven task categories, and eight frontier models. The best model reaches 52.8% accuracy. Better than spatial, but it means the best agent in the world still fails nearly half the time on routine analysis tasks.

scBench tests what bioinformaticians actually do

Existing biology benchmarks reward textbook recall or literature-style reasoning. Real analysis requires loading messy datasets, writing code, making judgment calls about thresholds and parameters, and producing concrete quantitative results. scBench recreates these conditions.

The benchmark covers 394 problems across six sequencing platforms (Chromium, BD Rhapsody, CSGenetics, Illumina, MissionBio, ParseBio) and seven task categories (QC, normalization, dimensionality reduction, clustering, cell typing, differential expression, and trajectory analysis). Each problem provides a data snapshot - typically an AnnData `.h5ad` file-captured right before a decision point, a natural-language task prompt, and a deterministic grader that scores the agent’s structured JSON output as pass or fail.

We pressure-tested each problem for shortcuts: removing precomputed embeddings, stripping cached labels, and ensuring answers can’t be guessed from prior knowledge alone. Cell typing (118 problems, 30%) and differential expression (71, 18%) dominate the benchmark because these are the stages where dataset-specific judgment matters most and where agents struggle hardest.

Agents are better at scRNA-seq than spatial - but still not reliable

Across eight frontier models from four providers, accuracy ranges from 29% to 53%.

Claude Opus 4.6 leads at 52.8%, followed by Opus 4.5 (49.9%), GPT-5.2 (45.2%), and Sonnet 4.5 (44.2%). The bottom tier - GPT-5.1 (37.9%), Grok-4.1 (35.6%), Grok-4 (33.9%), and Gemini 2.5 Pro (29.2%) - trails by a wide margin. The 23.6 percentage point spread between best and worst exceeds SpatialBench’s 18.3 point spread, meaning scBench discriminates model capability even at higher overall accuracy.

The accuracy bump over SpatialBench likely reflects training data. scRNA-seq has far more public datasets and Scanpy dominates the ecosystem with extensive documentation and tutorials.

Tasks reveal a consistent difficulty gradient

Not all analysis steps are created equal. Normalization is easiest (cross-model mean 70.4%), followed by QC (55.3%). These are procedural---apply a well-known transformation, check a metric against a threshold. Agents handle them reasonably well.

The story changes for judgment-heavy tasks. Clustering drops to 38.3%, cell typing to 34.9%, and differential expression to 27.0%. Seven of eight models follow this same difficulty ordering. DE is also where models diverge most: 27.7 percentage points separate best from worst.

The pattern is clear. Tasks that require contextual scientific reasoning-selecting marker genes, interpreting cluster identity, choosing statistical tests, identifying tissue-specific signatures-are where agents break down. General-purpose coding skill is necessary but not sufficient.

Platform choice matters as much as model choice

This is perhaps the most striking finding. Cross-model mean accuracy ranges from 59.1% on CSGenetics to 26.4% on MissionBio---a 32.7 point gap that exceeds the 23.6 point spread between best and worst models. Platform matters more than which model you use.

CSGenetics is easiest for six of eight models. MissionBio is hardest for all eight. The MissionBio collapse is dramatic: even the best model (Opus 4.6) only hits 42%, and Gemini drops to 10.3%. Every model shows large platform swings---Gemini drops 42 points between its best and worst platform. Even Opus 4.5, the most consistent model, loses 39 points.

MissionBio inverts the overall rankings in interesting ways. Grok-4 (sixth overall) beats GPT-5.2 (third overall) on MissionBio. Sonnet 4.5 surpasses GPT-5.2 by 11 points. Models that memorized Scanpy tutorials without learning transferable analysis techniques collapse on platforms with non-standard data structures and less common tooling.

These effects almost certainly reflect training data composition. Chromium and Illumina dominate public repositories and documentation. MissionBio and ParseBio appear less frequently. Reliable agents will need platform-aware context and assay-specific tooling, not one-size-fits-all reasoning.

scBench complements SpatialBench across single-cell modalities

Together, scBench and SpatialBench cover the two dominant transcriptional assays. The top model reaches 52.8% on scBench versus 38.4% on SpatialBench-scRNA-seq is more tractable. But the structural patterns are shared: normalization is easiest in both, platform effects drive 30--40 point swings in both, and model rankings are preserved at the extremes (Claude Opus leads both, Gemini ranks last in both).

The benchmarks form a complementary diagnostic. scBench tests whether models can handle the most common and best-documented assay type. SpatialBench tests whether that competence transfers to newer, less standardized technologies. Together they reveal whether an agent has learned general analysis reasoning or just memorized Scanpy workflows.

Limitations

Deterministic grading enables verifiable evaluation but necessarily discretizes scientific judgment into automatically checkable chunks. Each evaluation snapshots a single workflow step rather than capturing the long-horizon iteration where errors compound and thresholds are revisited. Real analysis is messier, more iterative, and longer than any single problem captures. But measuring step-level competence reliably is a prerequisite for automating longer workflows.

Where this is going

scBench confirms the pattern SpatialBench established: agents for biology are in a capability regime where they can accelerate routine analysis but cannot yet be trusted to autonomously answer scientific questions without human oversight. The path forward is a long tail of tractable engineering: platform-specific context, better harness design, and exposure to representative workflows across diverse biological contexts.

We’re building toward a family of benchmarks spanning major biological modalities, each an evolving formalization of the tacit knowledge and judgment that working scientists bring to data analysis. The goal is test-driven development of agent systems that improve through both model training and harness engineering.

Code, canonical evaluations, and full trajectories are available at github.com/latchbio/scbench.

This gap is starting to matter. Spatial assays (and increasingly every modern measurement modality) generate outputs that are too large, too irregular, and too context-dependent for most labs to analyze without computational specialists. The promise of agents is clear: point them at raw data, ask scientific questions in plain language, and compress weeks of analysis into hours. But looking right isn’t the same as being right - especially when results guide expensive and time intensive experiments.

Software engineering arguably went through this phase already. Frameworks like SWE-bench exposed how bad models were at real programming tasks and gave everyone a number to climb. Progress in biology needs something similar: built from real workflows, grounded in real datasets, and graded by verifiable outcomes instead of vibes.

Here we introduce SpatialBench (read the paper), a suite of spatial biology analysis problems designed to recreate the work scientists actually do - inspect messy images and count matrices, write and run code, reason scientifically and recover concrete biological results.

SpatialBench approximates real analysis work

Current biology evaluations largely reward textbook knowledge, but real analysis involves empirical interaction with data to produce answers. SpatialBench is a suite of 146 verifiable problems drawn from real spatial analysis workflows, spanning five platforms (Vizgen MERFISH, Takara Seeker, 10x Visium, 10x Xenium, Atlasxomics DBIT-seq) and seven task categories (QC, normalization, dimensionality reduction, clustering, cell typing, differential expression, and spatial analysis).

Each problem is a snapshot of an analysis workspace captured right before a decision point: typically an AnnData object plus any associated files, images, or outputs from earlier steps. The tests probe for durable biology, robust to survive reasonable choices of algorithm, library, or hyperparameters. We pressure-tested each for shortcuts: inspecting logs, adversarial prompts and removing anything solvable by pattern-matching or leaning on general scientific knowledge instead of actual analysis. Find concrete examples in the paper supplements.

Agents aren’t quite there yet

Across the full benchmark, base-model accuracy sits between 20-38% depending on the model family, with the usual tradeoffs between accuracy, latency, and cost.

But the aggregate numbers hide detail. Stratification reveals clear model-specific variation over tasks and platforms. Some model families collapse on specific task types while others swing 15 to 20 points depending on the class of spatial technology.

Spatial biology is of course a collection of many distinct technology types and constituent steps, each with its own artifacts and conventions obscured in point estimates of accuracy. Deliberate engineering will be needed to improve across each covariate.

Harnesses are more than “glue code”

Interestingly, harness design moved outcomes as much as swapping the base model. We ran Opus-4.5 under multiple execution ‘wrappers’ - LatchBio, Claude Code and a slightly modified mini-swe-agent - and saw dramatic variation.

Our results suggest we should start treating harnesses as first class objects in engineering and benchmarking efforts. The tools, prompts, and control flow around a model are as important as the model itself.

Agent trajectories reveal distinct problem solving strategies

Manual inspection of agent trajectories - the sequence of reasoning traces, tool calls, logs from an eval session - was perhaps the most informative part of the project. This data provides mechanistic explanations for performance differences. We could pick out distinct failure modes: models burning steps on formatting with no forward progress; using general QC thresholds that don’t hold for spatial data. Some models were able to ‘productively explore’ (inspect, compute, refine, make some forward progress) while others thrashed (same failed attempt on repeat, often forever). There are many examples of these in the paper.

Limitations

Verifiability comes with tradeoffs. Nuance is lost trying to draw deterministic lines around messy real world examples. Snapshots of single steps aren’t full workflows: real analysis is longer-horizon and iterative. But measuring step-level chunks with reliability is but a step towards automating truly long time horizon tasks. More on this soon.

Where this is all going

SpatialBench is a first contribution towards a broader family of benchmarks spanning major biological modalities. We view benchmarks as evolving formalizations of tacit scientist behavior and analysis workflows. They guide test-driven development of agent systems through model training and harness engineering that will eventually automate much of what we consider to be computational biology.

2026 will be a big year for agents in biology.

Read the paper. Check out data + code. Play with some examples.

Try it here: agent.bio

We’ll go over:

• Our roadmap for reliable and widely deployed spatial agents

• Important considerations while playing with these sandboxes

• Video demos of each agent

Our roadmap for reliable and widely deployed spatial agents

LatchBio develops agents for spatial biology data that allow scientists to direct their own analysis with natural language. By tailoring each agent to specific spatial kits and machines, and iterating relentlessly on this reduced scope, we believe it’s possible to build accurate systems that can be used to make expensive scientific decisions.

This release represents a shippable intermediary towards this goal, both to show technical progress and help others see why we’re so excited about this direction. The path ahead is a long tail of soluble engineering work: continued progress on benchmarks, purpose built infrastructure and technology specific heuristics encoded into prompts/evals.

Important considerations while playing with these sandboxes

We’re developing two agentic modes, “step-by-step” and “proactive”, that provide control over how often the agent will check-in, ask questions and provide figures or statistics for scientists drivers to sanity check intermediate steps. The sandboxes default to “proactive” to provide a nice first impression of end-to-end functionality, but we encourage biologists to start with “step-by-step” for any work of value. More on this soon.

Furthermore, the sandboxes were built for specific datasets and tasks. You can try to use your own data, and the system will likely generalize accurately on some tasks, but we provide no guarantees on performance quite yet.

Video Demos

The following videos and descriptions motivate some of the practical end-to-end workflows possible with the system. We highly encourage the reader to play with each themselves.

Takara Seeker

Ovulation failure affects over 10% of infertility cases in women of reproductive age in the United States, yet it is exceptionally difficult to study because ovulation unfolds within tiny, spatially restricted niches and evolves on the scale of minutes.

Using a Takara Seeker 3x3 kit, Mantri et. al built a spatiotemporal molecular atlas of the ovulating mouse ovary across 8 different time points.

To replicate the analysis, the Takara agent ingested raw H5AD files, removed off-tissue beads, computed QC metrics and filtering, performed clustering, and ran differential gene expression. The agent then manually reasoned through top marker genes for each cluster and identified cell types key to the ovulation process, such as mural and cumulus granulosa cells, mesenchymal subtypes, oocytes, and endothelial and immune populations. A scientists can then ask the agent to count the atretic or antral follicles.

Vizgen MERFISH

The mouse brain contains highly organized cellular circuits, yet these structures are difficult to resolve without single-molecule spatial data. MERFISH overcomes this barrier by capturing the precise locations of individual transcripts across intact tissue.

Using a MERFISH mouse brain dataset, the agent built a spatial neighbors graph, computed Moran’s I, and performed neighborhood enrichment and co-occurrence analyses. It identified spatially variable genes such as MYH11, SLC17A7, and DAAM2, each marking distinct anatomical regions; MYH11 was correctly enriched in blood vessel-associated cells, consistent with its vascular smooth muscle role. Neighborhood enrichment further recovered expected spatial organization, including co-localization of excitatory and GABAergic neurons and strong oligodendrocyte self-enrichment, matching known cortical and subcortical structure.

AtlasXomics DBiT-seq

Tumor heterogeneity is a major challenge in cancer, arising both between patients and within a single tumor. Changes in chromatin accessibility are a key driver of this variability because they alter gene expression, cell identity, and interactions in the tumor microenvironment.

Using AtlasXomics spatial ATAC-seq (DBiT-seq) from a stomach cancer sample and a matched control, the agent ran an ArchR-based workflow to identify differentially accessible chromatin regions, peaks, and motifs.

In tumor tissue, the agent detected strong enrichment of CTCF/CTCFL motifs, pointing to disrupted 3D chromatin organization, a hallmark of many cancers. It also found IRF1 enrichment, consistent with immune infiltration or localized inflammatory signaling. By contrast, the control tissue was enriched for FOXA1 and TEAD motifs, which support normal epithelial identity and Hippo pathway-mediated tissue homeostasis.

10X Xenium

Alzheimer’s disease reshapes the brain through localized inflammatory niches and region-specific neuronal vulnerability. These patterns only become visible only with high-resolution spatial transcriptomics.

Using a Xenium dataset from a TgCRND8 Alzheimer’s mouse brain section, the agent loaded the raw outputs, built an AnnData object, performed clustering and differential expression, and annotated cell types using a controlled mouse brain vocabulary.

The analysis recovered five major neural populations: astrocytes, excitatory neurons, inhibitory neurons, oligodendrocytes, and microglia. It also detected Alzheimer’s-relevant signals such as strong Gfap spatial enrichment marking reactive astrocytes and Hexb expression marking activated microglia. Finally, the workflow identified 21 spatial domains that correspond to known anatomical regions and disease-affected microenvironments, including glial clusters surrounding neurons, a characteristic pattern in Alzheimer’s models.

10X Visium

Colorectal cancer is shaped by tightly packed tumor, stromal, and immune niches that often sit only a few microns apart, making high-resolution spatial profiling essential for understanding disease architecture.

Using the Visium HD colorectal cancer dataset with more than 500,000 spots, the agent converted Space Ranger outputs into an H5AD and launched a GPU-accelerated RAPIDS workflow to handle the dataset at scale. The full pipeline from QC to clustering to differential expression completed in under a minute.

The agent then inspected top marker genes for each cluster and assigned epithelial, stromal, and immune cell types, reconstructing the organization of the colorectal tumor microenvironment.

Path Ahead

We’ve learned a lot about biology specific agent engineering in the past month building and saturating our benchmarks (eg. what is the analogue of a “code diff” for spatial analysis). Expect more educational material on these topics in the coming weeks.

High scores on SpatialBench represent expert and industrially relevant analysis ability across four different spatial technologies, tissue types and diseases. We benchmarked our spatial agent against Sonnet 4.5, GPT-5-Codex and Qwen 2.5 Coder in a minimal harness (a slightly adapted version of the Bash only SWE-bench harness). These early results show considerable engineering is needed to extend computer use agents to biology. There is a lot of work remaining to saturate these benchmarks.

We’re also releasing a batched test harness, eval spec and a representative subset of tasks/datasets on GitHub. While we built this tool to measure and improve our own product, we hope others find the open resources useful.

SpatialBench evals represent practical analysis tasks

This benchmark is focused on real world analysis work and each task was constructed actual chunk of work from our customers and their scientists.

Each eval is a JSON object conforming to a standard structure: a task description, data node and grader configuration. This information is injected into a system prompt and the agent performs work in a sandbox with access to data. A grader then evaluates the answer.

Here we demonstrate some concrete examples to build some intuition for what the benchmark is actually measuring. The specs for these evals are available in the repo.

Normalization

Feature Selection

Dimensionality Reduction

Clustering

Differential Gene Expression

Domain Detection

Results demonstrate need for biology specific agent engineering

While the benchmark does not represent an apples-to-apples comparison of systems, eg. a ~100 line mini-swe-agent harness wrapping Sonnet is significantly less complex than our system, the purpose of the exercise was to show what worked for coding will not necessarily work for biology. We will be improving the harness and expanding the eval bank over the coming weeks.

Putting a model in a simple loop with access to a shell is sufficient to saturate SWE-bench, but agents for spatial biology likely need purpose built infrastructure, tools and context for useful work. More on the details here.

Evals encode tacit knowledge of frontier research biology

The frontier of research biology - practical knowledge about details of different diseases and tissues - is distributed across hundreds of labs, in the minds of PIs and researchers. By working with these labs to interpret spatial data from different tech, we can encode their knowledge and preferences into evals used to train and benchmark agentic systems. These eval banks will become valuable, structured documentation about the procedural details of research as they grow.

Add your spatial technology to our benchmarks

We are expanding SpatialBench to include all relevant machine and kit types. Reach out to our team to include your technology in our next release.

Unlike most experiments in biology, the computational steps in spatial can match or exceed the resources spent on upstream work in the wet lab: a tricky combination of the data being large (>1TB a run for CosMX) but the analysis high-touch, multi-turn and very unique to each scientists’ questions. This is unfortunate because most biologists have a very clear idea of what they are trying to do, often able to describe the end results/plots with great detail and steer intermediate steps based on their biological goals. But the practical complexity of using spatial tools complicates the translation of the correct ideas to the right analysis. Tens of thousands of scientists across industry and academia are stuck analyzing spatial data .

Agents are LLMs in loops that plan and execute end-to-end analysis with natural language directions by interacting with their environment and calling tools. They have shown incredible promise in math and programming, but require intentional engineering to translate well to biological data. Latch spatial agents are tailored to the details of each kit, in direct collaboration with the companies that sell them, so they have the specific context, tools and structured analysis flows required for each type of data. They build scientific notebooks with interactive widgets running on large machines and launch workflows on purpose built data infrastructure engineered for scale.

We are in the early days of this emerging category, but the initial technical results and response from scientists indicate this is likely the future of data analysis. There is considerable work left to do, but it seems base reasoning and agentic capabilities of frontier models are essentially here, and the remaining structure of the problem is a long tail of incremental but very soluble engineering improvements. In the coming months, tens of thousands of scientists will be able to finally play with their own data and answer their own questions.

• Agents for biological data are purpose built for the domain

• Spatial data is a strong focus for early adoption

• What do agentic workflows look like in practice?

  • AtlasXOmics DBiT-seq Agent

  • Takara Trekker Agent

  • Vizgen MERSCOPE Agent

  • 10X Xenium Agent

Agents for biological data are purpose built for the domain

Agents need new capabilities to reliably and accurately accomplish practical tasks with biological data. Unlike software engineering products, these systems need to run in sandboxes with considerable computing and storage resources to manipulate large data. They also need to offload particularly intensive steps to optimized workflow infrastructure; single operations, like processing sequencing reads or clustering counts, can take hours and require clusters of computers. Using pre-existing bioinformatics components as tool calls, like Nextflow or Snakemake workflows, also serves to constrain end-to-end workflow variability and allows models to focus on genuinely novel components of user requests.

Furthermore, the driver of the agent is a scientist who reasons about results with plots, charts and graphical components. Latch agents communicate with their users by building scientific notebooks, rich with markdown description, that also serve as self-documenting dashboards that describe the analysis in detail.

Perhaps most important, each agent is tailored to the specific type of machine or kit that generated the data. We work directly with the companies developing these assays when possible - Vizgen, TakaraBio, AtlasXOmics - to encode latent knowledge about the specific details, hidden footguns and ideal customer journey into prompts, tools and evals. Rather than an open-ended sandbox that does many things somewhat well, our philosophy is to build a surgical and highly reliable product that assay providers can trust. We believe the bar for widespread adoption of agents is an intelligent, adaptable system that does not jettison the determinism of traditional software.

Spatial data is a strong focus for early adoption

It is important to find specific tasks that are uniquely suited to agentic workflows and cause enough existing pain to encourage scientists to try a new tool. Spatial data checks both of these boxes. This is a new, rapidly growing but notoriously challenging domain of science. There are no clear analysis standards, the ecosystem of tools is highly fragmented and generally difficult to use, and unlike traditional sequencing, there are many different, composable steps. Yet the power of these assays is remarkable, in the new biology they help us understand and the beautifully intuitive way they allow scientists to see what’s going on.

In many ways, it is convenient agentic systems are beginning to actually work. It is unclear exactly what the solution to these problems would have been with traditional software: likely many variations of hard coded dashboards that each tackle a specific task in a reliable but brittle manner. We believe the interplay between spatial biology, which is going through its own revolution, and agentic data analysis is a special, reciprocal relationship. These complementing technologies will allow scientists to ask more, understand more and all the while consume more kits and compute.

What do agentic workflows look like in practice?

The following are four concrete end-to-end agentic workflows built in collaboration with real scientists. A tailored agent is built for each kit or machine type. Scientists begin with raw data, send instructions or questions to a chat interface (eg. “Process my Vizgen data”), and watch the agent build graphical scientific notebooks to accomplish tasks.

AtlasXOmics DBiT-seq Agent

AtlasXomics develops an assay called deterministic barcoding in tissue sequencing (DBiT-seq) that allows spatially-resolved epigenome profiling. Microfluidic chips deliver barcodes to precise tissue coordinates, uncovering relationships between chromatin accessibility, histone modifications, and context-specific gene expression within intact tissues.

This first flow highlights the design principles discussed earlier: large agent sandboxes to handle the TBs of file data, the use of existing bioinformatics workflows developed by the AtlasXOmics team and specialized visualization components (H5, IGV viewer) required for this data type.

1/ Bioinformatics Workflows as Tool Calls

The AtlasXOmics agent begins by building interfaces to handle user input and dispatching bioinformatics workflows to purpose built data infrastructure. Reads are filtered, aligned to the reference genome, and converted into fragment files.

It understands the relevant files to pull from the output folder - filtered fragments, spatial position files, motif enrichment scores and gene activity scores - so scientists can immediately continue with downstream steps. Scientists can ask questions about workflow parameters at any point.

2/ Image Alignment

Linking molecular measurements to precise tissue locations requires careful alignment of sequencing data with high-resolution tissue images.

The agent guides scientists through this process using an interactive interface, providing instructions on how to perform alignment and suggesting adjustments when needed. The result is a spatially-aligned dataset where each molecular measurement is mapped to its exact tissue coordinate, creating the foundation for all downstream spatial analysis.

3/ Spatial Region Selection

Once the data is aligned, scientists need to define regions of interest based on tissue morphology or biological features. The agent enables researchers to use intuitive lasso-select tools to define custom regions

It can suggest regions based on expression patterns or guide selection based on natural language descriptions like “select the tumor border region.” These defined spatial regions are saved as metadata, ready for comparative analysis across different tissue areas.

4/ Differential Chromatin Analysis

Understanding regulatory differences between conditions, spatial niches, or cell types requires identifying differential chromatin accessibility regions and motifs. When scientists ask questions like “What chromatin regions are more accessible in the tumor core versus periphery?”

The agent launches the appropriate differential analysis workflow as another bioinformatics workflow, automatically navigating the Latch file system using the LPath API to find required inputs, and configures parameters based on the scientific question. The analysis produces differential peak results, motif enrichment comparisons, volcano plots, heatmaps, and gene regulatory network differences.

5/ Subclustering

Discovering subtle cell populations or states within major cell types often reveals distinct regulatory profiles crucial for understanding biology. When scientists want to explore specific cell populations in detail, the agent subsets the data based on user criteria and launches re-clustering workflows on the subset.

It then repeats annotation and differential analysis on subclusters while maintaining the full analysis provenance in the Latch Plots notebook environment. This iterative process yields refined cluster definitions and subcluster-specific regulatory signatures.

6/ Visualizing Chromatin Accessibility Tracks with IGV Browser

Understanding how chromatin accessibility differs between clusters requires examining the actual peaks and signal intensities at specific genomic loci. When scientists want to validate differential accessibility results or explore regulatory elements in detail, the agent launches an IGV browser session and automatically loads cluster-specific chromatin accessibility tracks as bigWig files, with each cluster displayed in distinct colors for easy comparison.

Scientists can navigate to genes of interest by simply asking “Show me chromatin accessibility at the IL2 locus across T cell clusters,” and the agent configures IGV to display the relevant tracks, adjusts the viewing window, and highlights differential peaks. The agent can also overlay additional tracks like gene annotations, known enhancers, or transcription factor binding sites, providing the full genomic context needed to interpret spatial epigenetic differences.

Takara Trekker Agent

TakaraBio offers two spatial transcriptomics platforms. This flow focuses on Trekker, a single nuclei spatial transcriptomics kit.

We show the agent’s ability to gather experimental context, launch chemistry-aware bioinformatics workflows, guide scientists through tissue image alignment, and discover insights through multi-sample differential gene expression analysis.

1/ Gather Experimental Context

The chemistry of each assay dictates the structure and content of its sequencing output and therefore the architecture of its downstream computational analysis. To enable robust interpretation, we codified assay-specific scientific logic into our agentic framework, allowing our Latch agent to reason about experimental design, platform chemistry, and expected data modalities.

The agent is trained to ask clarifying questions about the kit chemistry, sample type, and sequencing platform before constructing the appropriate workflow. This ensures that each dataset is processed using a computational path aligned to its biochemical origin, enabling accurate spatial reconstruction and downstream biological analysis.

2/ Chemistry-specific Bioinformatics Workflows

The workflow ingests paired FASTQ files alongside output from a single-cell platform (e.g., 10x or BD Rhapsody). Each read encodes a cell barcode, molecular identifier, and spatial barcode. The Takara spatial informatics pipeline identifies nuclei carrying valid spatial barcodes, reconstructs the spatial index, and generates a cell × gene matrix that integrates molecular expression with spatial localization.

When prompted with, “Process FASTQ files from a Takara Trekker experiment,” the agent dynamically generates a configuration interface to capture parameters such as the single-cell platform, kit version, and output format. It then invokes the appropriate Latch `w_workflow`, which executes on dedicated high-performance infrastructure to produce an H5AD object ready for visualization in the Latch plotting environment.

2/ Image Alignment and Tissue Visualization

Seeing gene expression overlaid on tissue structure is crucial for understanding spatial data. However, aligning H&E-stained images with spatial coordinates is challenging and often requires manual tweaking. The agent simplifies this using built-in image registration tools in the H5 viewer.

When prompted, the agent automatically focuses on the tissue region and overlays the H&E image with the spatial data. For ovarian tissue, this alignment reveals how gene expression maps to anatomical features: follicles appear as distinct circular structures, corpora lutea show as larger irregular regions, and the surrounding stroma displays its own expression patterns. This visual overlay lets researchers immediately connect molecular data with tissue architecture.

3/ Differential Gene Expression Across Follicular Subtypes

We also demonstrated the agent’s ability to conduct differential gene expression analyses across ovarian follicle subtypes—preantral, antral, and atretic—within Seeker-derived datasets. This analysis established a spatially resolved baseline for folliculogenesis, linking molecular identity to anatomical localization, revealing spatial gradients in follicle maturation, and identifying novel, spatially restricted gene markers critical to understanding ovarian development.

When extended to temporal dynamics, the agent identified stage-dependent changes in gene expression following hCG induction, including a decline in Fabp3 and Inhbb levels over time, marking the transition of mural granulosa cells from a proliferative, follicle-maintaining phenotype to a luteinizing, pre-ovulatory state.

Vizgen MERSCOPE Agent

Vizgen’s MERSCOPE platform uses multiplexed error-robust fluorescence in situ hybridization (MERFISH) to map hundreds to thousands of RNA species simultaneously at subcellular resolution. Each experiment generates terabytes of data: millions of transcript coordinates, multi-channel z-stack images, and complex metadata files.

We show the agent’s ability to process TB scale MERFISH human brain imaging data, launch GPU-accelerated cell segmentation workflows, enable interactive visualization of millions of transcripts in the browser, and perform spatially-aware differential expression analysis through natural language commands.

1/ Cell Segmentation

Cell segmentation determines how millions of detected transcripts get assigned to individual cells. MERSCOPE captures five imaging channels across seven z-planes: DAPI (nuclei), three Cellbound channels (membranes), and PolyT (cytoplasm). Creating accurate cell boundaries requires selecting the right channel combinations and configuring GPU-accelerated algorithms like CellPose or Baysor.

The agent automates this complexity by scanning imaging directories and generating configuration files tailored to each tissue type. For dense tumor regions, it prioritizes membrane channels; for sparse samples, it leverages Baysor’s transcript-based approach. It then launches a GPU-powered workflow on Latch infrastructure to process the segmentation.

The workflow outputs three files: (1) an AnnData object compatible with downstream tools like Scanpy and Squidpy, (2) PMTiles (a pyramidal format originally developed for web mapping that enabling smooth zooming and panning), and (3) a DuckDB database containing transcript coordinates and metadata, providing a lightweight SQL engine for querying millions of transcripts in real-time.

These files are displayed in the Latch H5 viewer, where scientists can seamlessly overlay cells, transcripts, and high-resolution morphology. They can also filter transcripts by gene, spatial region, expression level, or cell type while exploring tissue architecture at any zoom level.

If segmentation needs refinement, scientists can immediately instruct the agent to rerun with adjusted parameters, creating a rapid iteration cycle for optimal results.

A video of agent performing all analysis steps end-to-end for Vizgen data, from QC, normalization, finding HVGs, to cell type annotation and differential gene expression

2/ Quality Control

With cells defined, quality assessment becomes critical. MERSCOPE’s large imaging area can harbor technical artifacts: edge effects, tissue folds, or regions with poor probe penetration that must be identified before analysis. The agent generates spatial QC plots revealing these issues, allowing scientists to set informed filtering thresholds that preserve biological signals while removing technical noise.

3/ Clustering and Cell Type Annotation

MERFISH panels contain 200-1000 genes carefully selected to distinguish cell types in specific tissues. The agent exploits this focused design, performing clustering at resolutions that match expected cellular diversity. Initial annotation queries reference databases for marker patterns, then scientists refine through natural language: “split the myeloid cluster into monocytes and macrophages” or “identify proliferating cells across all types.” Spatial context often resolves ambiguous populations. Morphologically similar cells may occupy distinct tissue compartments that clarify their identity.

4/ Differential Expression Analysis

With cell types defined, scientists often ask what makes populations distinct or how they change across conditions. When comparing “tumor cells in the core versus edge,” it accounts for spatial autocorrelation and multiple testing across the focused gene set. Results highlight not just which genes differ, but their spatial patterns: gradients of metabolic genes from hypoxic cores to vascularized edges, or immune activation markers concentrated at tumor-stroma interfaces.

10X Xenium Agent

10X Genomics’ Xenium platform uses in situ hybridization to detect hundreds of RNA species at subcellular resolution. To demonstrate the agent’s capabilities, we’ll analyze an Alzheimer’s disease mouse model dataset tracking pathological changes across disease progression.

We show the agent’s ability to parse complex multi-timepoint Xenium datasets, refine cell segmentation in challenging brain tissue, detect spatial domains through BANKSY clustering, and calculate disease-specific gene signatures that reveal spatially-organized neuroinflammatory responses around amyloid plaques.

1/ Loading and Merging Xenium Data

The dataset compares wild-type and CRND8 transgenic mice at timepoints representing mild, moderate, and advanced amyloid deposition. Scientists attach Xenium outputs from coronal brain sections profiled with 347 genes: 248 from the Mouse Brain panel plus 99 custom genes targeting activated microglia, astrocytes, and plaque-inducible markers.

The agent parses the complex directory structure, constructs an AnnData object preserving spatial coordinates, and generates diagnostic plots.

2/ Cell Segmentation Refinement

Brain tissue presents unique segmentation challenges. Neurons range from tiny interneurons to large pyramidal cells spanning hundreds of microns. Around amyloid plaques, activated microglia cluster densely, creating cellular aggregates that standard segmentation often merges incorrectly.

When prompted to “improve segmentation around microglial clusters,” the agent launches a refined workflow. Cellpose first identifies nuclei using DAPI, providing reliable seeds even in crowded regions. Baysor then expands these boundaries guided by microglial transcripts like Iba1 and Cx3cr1. This approach captures the complex morphology of activated microglia wrapped around plaques–essential for accurately quantifying neuroinflammatory responses across disease stages.

3/ Spatial Domain Detection

With proper segmentation complete, the agent reveals tissue organization through BANKSY clustering.

4/ Cell Type Annotation

Standard cell typing identifies expected populations, but the key insights come from disease-specific signatures. When scientists prompt to “compute PIG signature scores within microglia and astrocytes,” the agent executes a targeted analysis of plaque-induced genes, a curated set of markers that distinguish reactive from homeostatic glia in Alzheimer’s pathology.

The agent first identifies microglia and astrocytes using canonical markers, then calculates PIG scores incorporating genes like Apoe, Cst7, and Axl for microglia, and Gfap, Vim, and Clu for astrocytes. These scores are mapped spatially, revealing striking patterns: microglia closest to plaques show the highest PIG scores, forming a reactive barrier, while astrocytes display elevated scores in broader halos around pathological sites.

Try it

Shoot us an email at contact[at]latch.bio to get onboarded. Lots of work to be done, but we have certainly seen early signs of the magic. We believe this will soon be the standard way biologists interact with data.

Every biotech executive has the same nightmare: a data breach halts your trial, exposes your critical IP, and scares off future investors. GDPR exists to prevent this nightmare; in short, it is Europe’s strict privacy rulebook (but, unlike most lab SOPs, this one comes with fines up to €20 million when things go wrong).1

Despite its reputation in biotech as a bureaucratic hurdle, GDPR reflects principles that not only protect data, but in practice, handle it to an end that supports the reproducibility and validation of experimental output for clinical translation.

For solution providers, GDPR’s value is clearer; it presents a commercial advantage and aligns your analysis with the same values that patients, clinicians, regulators, and software solutions already demand: accuracy, transparency, and accountability. It creates a common standard upon which geographically-diverse client data moves seamlessly from research to clinic, across international borders, and through tools without losing integrity along the way.

This post will apply key GDPR principles to the lab, operationalizing them with LatchBio’s approach to default-compliant product development.

First, an Overview of Compliance at LatchBio

LatchBio’s Global Compliance Program is designed to optimize the speed of your discovery without sacrificing its security. We work closely with both technical and regulatory partners to ensure the highest level of compliance and confidence for our customers:

• Drata provides 24/7 compliance monitoring visible to you,

• NAMSA advises on biotech-specific regulatory positioning,

• Tenax Solutions conducts annual penetration testing,

• MJD Advisors audits the integrity of our commitments each year, and

• GRCI Law employs our EU- and UK-based GDPR representatives and advises on international transfers.

Our aim is to meet every client’s regulatory requirements and work closely to provide regulatory support needed from a data processor hosting critical scientific infrastructure. To this end, we are compliant with a range of US and international frameworks, and quickly work with our customers when they need custom requirements or to comply with a different regulatory framework.

Latch is compliant with HIPAA, SOC 2, GxP, CLIA, CCPA/CPRA, and more. Visit our Trust Center to view our extensive security and compliance documentation.

Data Provenance and Reproducibility for Patient Rights

Across industries and countries, the public increasingly demands visibility into what companies know about them (e.g., requesting a copy of your personal data from Meta, or asking 23andMe to permanently delete your data). GDPR codified this cultural shift into law by giving patients clear rights to access, correct, or delete what companies know about them,2 including biotechs and solution providers.

GDPR and complementary frameworks have requirements that are satisfied by immutable logs of processing history (e.g., every workflow, data transfer, analysis step, or single data manipulation),3 with evidence of who did what and when. If done poorly, this could result in a disarray of paperwork and messy file systems (think graduate student’s desktop with folders inside folders, labeled “final_v7_actually_final).”

Latch’s architecture treats auditability and provenance as defaults. The full history of manipulation and record is available without manual reconstruction. The result is a system aligned with GDPR’s principles of accountability and transparency.4

The value of this requirement and solution can be interpreted beyond data subject rights. It allows scientific teams to reproduce an experiment exactly as it was run, returning to prior states and tracing how data evolved over time without guesswork to confirm the reliability of results.

Principle of Least Privilege in Modern Bioinformatics Management

Research data moves constantly, between collaborators, instruments, and workflows. Each transfer introduces potential for error: a file overwritten by someone who didn’t need editor access, a dataset manipulated without context, a result tied to the wrong version of an input. GDPR formalizes what lab operational excellence already demands — that access to data is deliberate, limited, and accountable.5

The principle of least privilege6 is, at its core, a safeguard against that kind of entropy. It requires admins to limit access to only what each user needs for their role and to keep detailed records of those interactions. In other words, it encodes good lab hygiene into law.

On Latch, this takes form through granular role-based access control (RBAC). Workspace owners define broad roles (viewer, editor, or admin) and then granularly customize permissions within those roles. Before high-impact actions like data export or deletion, user identity is re-verified7, and Latch engineers would have already been monitoring automatic signals for suspicious log-in behavior to report to admins.

The benefit isn’t just compliance; it’s coherence. Least-privilege access enforces experimental discipline: teams work from a single, trusted source, data integrity maintains intact, and the UI of each employee is specifically designated by the admin. What GDPR describes as “security of the processing” is, in practice, the same discipline required for reproducible, multi-stakeholder science.8

Commercial Advantage of GDPR Compliance for Biotech Solution Providers

In any pharma sale, the client’s procurement team will ask for a key set of GDPR-required artifacts, to size up a vendor’s readiness for regulated data transfer:

• Standard Contractual Clauses (SCCs) define the terms for international data transfers9,

• Data Processing Agreement (DPA) outlines how a processor manages and protects a controller’s data10,

• Record of Processing Activities (ROPA) documents what data is handled, by whom, and for what purpose11,

• Data Protection Impact Assessment (DPIA) evaluates risks in higher-impact processing, such as clinical or genomic data12,

• Transfer Impact Assessment (TIA) assesses legal and technical safeguards for cross-border data13.

For solution providers, familiarity and preparation with these isn’t bureaucracy; it’s evidence of comprehensive risk management and sufficient understanding of complex international compliance. It shortens procurement cycles, and removes one of the most common blockers to large-scale deployments. Teams that can present these materials quickly are viewed as reliable partners who understand both the regulatory and operational realities of handling sensitive patient-derived data.

Latch integrates these same requirements directly into its onboarding and platform architecture. SCCs, DPAs, and supporting assessments are maintained and reviewed with independent auditors, creating a security and documentation layer that travels with your work. This foundation allows solution providers to engage with regulated customers without delay, building credibility while keeping the focus on scientific delivery rather than administrative process.

In Conclusion: Global Data, AI Models, and the New Relevance of GDPR

As science globalizes, GDPR’s reach extends naturally with it. Sequencing data, pathology images, and cellular atlases now cross continents daily to train AI models or power multi-omics pipelines. These datasets almost always contain information traceable to European individuals (directly or indirectly) which places them squarely under GDPR’s jurisdiction.

For solution providers, this means that data sovereignty and privacy controls are now core to scientific scalability. When you train an AI model on European datasets, deploy compute in the U.S., or serve results to a collaborator in Asia, GDPR follows the data at every step.

Far from being an obstacle, this consistency enables collaboration. Standardized privacy frameworks like GDPR create a shared security language between institutions, reducing friction in multinational research.14 They let scientists exchange data and model outputs with confidence that each participant is adhering to compatible privacy and accountability standards.

In this sense, GDPR has become part of the infrastructure of global bioinformatics: a framework that allows sensitive patient-derived data to move across borders, feed algorithms, and return as clinically meaningful insight, all while preserving the trust that underpins scientific progress.

If you are a solution provider interested in white-labeling these tools and offering them to customers through a branded portal, email contact@latch.bio.

LatchBio’s Compliance Team is available at compliance@latch.bio.

• Overview

• Deploying Forch

  • Task IPv6

  • Task Firewalls

  • Node Mounts

  • Task Configuration

  • Task Secrets

  • Node Autoscaling

  • Public IPs

  • Storage Devices

  • Cuda Devices

  • Billing

• Architecture Overview

• Using Forch

Overview

Forch has native support for cross-region (and even cross-cloud) deployments. This brings improved network latency for Latch Pods/Plots and other interactive software for users outside the US west coast, where our compute is currently hosted. It also gives us access to a much larger pool of instances - when the AWS datacenter in Oregon runs out of A100 GPUs, we can opt to get some from GCP in Utah. We can even choose the location by comparing prices of all suitable instances across all regions of all providers, bringing a potentially significant reduction in cost.

It should also be possible to eventually support on-prem deployments as it doesn't make a difference to the orchestrator where the computer is located. Forch is designed to avoid up-front requirements for its compute nodes and their environment. For example, if you never run a task which wants to accept internet connections, you never have to worry about the public IP allocation system or firewall configuration. The Forch components responsible for these features will never run. If the on-prem nodes happen to already have storage devices attached to them, Forch tasks can use the devices without a storage device driver in the orchestrator itself. This means we can automatically support environments without cloud APIs.

As opposed to our Kubernetes setup, Forch does not choose from across pre-defined "node groups". Instead it queries the cloud providers directly for all available instance types and chooses according to a scoring formula. This means that not only is the cost reduced by choosing the best location, even the type of the instance is chosen based on price. Moreoever, Forch uses a generic constraint solver to decide which tasks will run on which nodes, allowing for for very complicated requirements and preferences. For example, you could encode data egress costs (based on estimated task input size) and include them in the price estimate for a given instance. In our Kubernetes set up (if it supported multiple clouds) you would have to manually check the instance prices, estimate egress costs, and make a decision ahead of time. By the time the task was running, the instance prices or availability might also be different, especially when using preemtible spot instances.

We also designed Forch to be easy to debug and maintain. Our experience running Kubernetes has been plagued by errors that are specific to the interaction between various semi-compatible versions of its plugins and the idiosyncracies of AWS EKS (our provider), the Kubernetes resource abstraction, and Go (its implementation language). Commonly, the only error exposed by Kubernetes APIs is the dreaded "context deadline exceeded" which means that something somewhere timed out and we have to scan through millions of lines of logs across dozens of services to find relevant information. Forch is built around a central database which stores the entire history of state transitions for all objects, which are themselves immutable. It is always possible to figure out how the state of the system evolved over time to get to an error. We also include OpenTelemetry-based distributed tracing in all our services to make sure we have very complete, structured information about the actions that Forch takes at the code level.

Deploying Forch

We are working on making Forch as easy to deploy in a third-party cloud account as possible. The current setup involves provisioning an AWS CloudFormation template which creates a setup role, which is passed to a Pulumi deployment to configure a dedicated VPC and other downstream roles. After the initial setup, the CloudFormation role can be optionally revoked as the Forch will use fine-grained roles managed by Pulumi. We expect to provide more flexible deployment options (using an existing VPC, or without any custom AWS IAM permissions) in the future.

In its current form, Forch assumes an IAM role in the target AWS account to issue API calls. We plan to eventually introduce a security barrier in the form of an "Agent" service which runs in the target account itself and acts on behalf of the orchestrator. This service will be fully controlled by the account owner and Forch will no longer require users to grant any permissions to the Latch AWS account.

We currently expect users to use a Latch-provided AWS EC2 AMI for the compute nodes though in practice there are only two requirements for a computer to join the cluster:

1. a Python runtime (typically with uv),

2. an internet connection of some sort (egress only being acceptable).

All features that require cloud APIs are designed to be optional as they may not be necessary for some users and require additional setup, account permissions, and potentially impose limitations on the compute environment. Forch currently supports the following:

Task IPv6

Forch will configure IPv6 addresses for each task if the node's network card reports an assigned IPv6 prefix. This is used by Latch Pods.

Task Firewalls

Forch configures firewalls using nftables on each node to prevent communication between the tasks, the tasks and local networks, and the tasks and the internet. These are fully configurable and allow rules based on specific IPs and task IDs (which resolve to the tasks' current IPv6 addresses). This is used by Latch Pods.

Example configuration (database rows):

task_id=1 peer_task_id=2 direction='ingress'

-- allow task #1 to receive traffic from task #2

Node Mounts

Forch allows tasks to request directories from the node are bind-mounted into the task. This is currently only available to "privileged" tasks. You may request privileged tasks only on nodes for which you have additional authorization explicitly for this feature. Forch currently uses this internally to implement NFS server tasks.

Example configuration (database rows):

-- note: the task must have privileged=true

task_id=1 src_path='/dev' mount_path='/dev' mount_propagation='rshared'

-- allow the task to access the parent node's /dev directory

Task Configuration

Forch allows passing configuration data to tasks in environment variables or injected files. The data is stored in plain text in the Latch-owned central database so is not suitable for secrets or sensitive information. This is used for all Latch workloads.

Example configuration (database rows):

-- task_datum_spec:

id=1 dest_environment_variable='TEST' src_inline='hello world'

id=2 dest_path='/config.json' src_inline='{"hello": "world"}'

-- task_data:

task_id=1 datum_spec_id=1

-- task #1 receives data from spec #1 (TEST='hello world')

task_id=1 datum_spec_id=2

-- task #1 receives data from spec #2 (file /config.json)

Task Secrets

Forch can use external secrets management APIs to securely pass secrets to the tasks in environment variables or injected files. The secrets are absolutely never read outside the compute node running the task requiring the secret. This is used for all Latch Pods.

We provide an AWS SecretsManager driver. This feature requires respective AWS permissions.

Example configuration (database rows):

-- secret:

id=1 provider_id='secret-key' infra_provider='aws' infra_account='000000000000'

-- task_datum_spec:

id=1 dest_environment_variable='KEY' src_secret_id=1

-- task_data:

task_id=1 datum_spec_id=1

-- task #1 receives data from spec #1 (KEY=<REDACTED>)

Node Autoscaling

Forch is able to determine when new nodes are required to run tasks and automatically request new nodes. It is also able to detect empty nodes and request their shutdown after a cool-down period. This is used for all Latch workloads.

We provide an AWS EC2 driver. This feature requires respective AWS permissions.

Forch also allows specifying a schedule anticipating expected demand by day+hour and Forch will scale up nodes in advance to ensure demand can be fulfilled.

Public IPs

Forch can assign public IPs to tasks, if they require routing for incoming connections. We use this internally to implement Latch Pods.

We support nodes with public IPv6 and IPv4 addresses (any number of assigned addresses would work).

We also provide an AWS Elastic IP driver. Forch also provides a custom NAT server to support using Elastic IPs with nodes from a private VPC subnet (i.e. the nodes do not have a pre-assigned public IP). This feature requires respective AWS permissions.

Example configuration (database rows):

-- ip:

id=1 provider_id='eipalloc-00000000000000000' infra_provider='aws' infra_account='000000000000' infra_geo_partition='us-west-2'

-- task_ip:

task_id=1 ip_id=1

-- task #1 will have IP address #1 (provided by AWS Elastic IP with allocation `eipalloc-00000000000000000`)

Storage Devices

Forch allows tasks to mount external devices for persistent storage. This is used in Latch Pods. For ephemeral storage you can use the node's storage devices. This is used for Latch Pods.

We do not currently support placing ephemeral storage on dedicated volumes, but this may change in the future.

We provide an AWS EBS driver which additionally supports volume resizing. We plan to eventually provide volume pre-warming and cross-region/AZ copying. You must create the volumes externally and provide the volume ID to Forch, but this is typically a single API request. This feature requires respective AWS permissions.

Example configuration (database rows):

-- storage_device:

id=1 provider_id='vol-00000000000000000' infra_provider='aws' infra_account='000000000000' infra_geo_partition='us-west-2'

-- task_storage_device:

task_id=1 dev_id=1 mount_path='/mnt' mount_type='filesystem'

-- task #1 will have device #1 (EBS volume `vol-00000000000000000`) mounted to /mnt

-- the node will ensure a filesystem is present on the device before the task starts

Cuda Devices

Forch nodes can attach CUDA devices to tasks, when available. They will also mount the node's CUDA drivers/SDKs into the tasks since the driver version has to match exactly between a container and the host.

Example configuration (database rows):

-- task:

display_name='test' ... dedicated_gpu_type='A10G' dedicated_gpu_count=1

-- task "test" will only run on nodes where an A10G is available

-- the device will be provided to the container along with the CUDA driver

-- using bind mounds from the host

Billing

Forch allows tracking the runtime of tasks and "charging" it to an entity by ID. The task creator must provide authorization to charge to a specific entity. This is used internally by Latch to implement billing.

Architecture Overview

The core components of Forch are:

1. The database (Postgres)

2. The orchestration service

3. The node daemons

4. The NAT servers (per VPC/networking environment)

The orchestrator handles scheduling using constraint solving and executes various environment control API calls e.g. attaching/detaching AWS EBS devices.

In the future, we plan to add an additional "Agent" service which will handle cloud API requests and other environment configuration functions which are currently assigned to the orchestration service. This is to improve permission separation by allowing Forch to function without holding permissions delegated from the compute environment e.g. with an externally-assumable AWS IAM role.

The node daemons send requests to the local container runtime based on tasks read from the database. They also handle system-level setup like firewall configuration (using nftables), storage device preparation (partitioning, file system creation for new devices).

All the services communicate exclusively through polling the database for new rows. Authorization is implemented in-database using row-level security functions, predicated on ED25519 JWKs verified by a plv8-packaged JS implementation. There is no API server, though Latch uses its existing API servers for creating workloads and workload resources (i.e. operating the Forch cluster). User-owned deployments will not need to interact with these services.

The NAT server works by setting up IP-IP tunnels to each node and configuring nftables. This is generally optional, depending on the networking setup. It was originally introduced because the AWS Elastic NAT service cannot be used with Elastic IPs since none of the available Route Table configurations route traffic appropriately. It should be rarely necessary outside of this setup.

Our database schema is based around immutable objects holding configuration for various resources (like provider-specific IDs e.g. AWS EBS volume IDs), and the associated event tables, which hold a parent object ID, timestamp, and type + their event data tables, which hold the data specific to each event type. The state of each object (e.g. if a storage device is attached to a node) is determined by taking the latest event and looking at its type and/or data. We hold the entire history of all state transitions of all objects for observability/debugging purposes.

All services are also traced using OpenTelemetry/Datadog. This is a huge improvement over the infamous Kubernetes/Go "context deadline exceeded" set of errors.

Using Forch

Latch users can look forward to Forch being enabled globally in 2026. For the time being, Forch is already available with explicit opt-in. Users can choose one of several Latch-owned deployments in various AWS regions and we will bring up more regions and other cloud providers over time based on demand. We are working with select partners to make user-owned cloud compute environments generally available by the time Forch rolls out to the public.

If you are interested in Forch, please reach out to contact@latch.bio or on social media.

We may publish detailed posts focused on the engineering behind Forch in the future. Please reach out to kenny@latch.bio or on social media if you are interested so we can prioritize topics accordingly.

Yet understanding the complete state of a cell requires more than measuring RNA. Gene expression is the outcome of a complex regulatory program, influenced by chromatin accessibility, histone modifications, and DNA methylation. Capturing these modalities in their native spatial context is essential for building a more complete atlas of biology.

Go to Data Portal

Today, we are releasing a new data portal with AtlasXomics dedicated to spatial epigenomics. The portal includes datasets from four publications, spanning diseases such as Alzheimer’s and gastric adenocarcinoma, along with studies of heart and brain tissue development.

These resources bring a new layer of information into focus, revealing not only which genes are expressed, but also how the regulatory architecture of the genome shapes that expression in space.

Visit the portal and study these datasets here: https://atlasxomics.latch.bio/data-portal

How does it work?

At the heart of these datasets is a technique called Deterministic Barcoding in Tissue sequencing (DBiT-seq), pioneered by AtlasXOmics. DBiT-seq uses microfluidic chips to flow barcoded oligonucleotides across tissue sections. Each intersection of these flows marks a defined spatial coordinate, which can then be linked to sequencing reads.

The result is a generalizable platform that can add spatial context to a variety of assays. When combined with ATAC-seq, for example, DBiT-seq produces spatially resolved maps of chromatin accessibility. Similarly, paired with CUT&Tag, it reveals the distribution of histone modifications such as H3K27ac or H3K4me3 across tissue.

What does the data tell us?

To understand what insights are revealed through DBiT-seq, let’s walk through an example dataset of how spatial ATAC-seq helps us better understand tumor heterogeneity and chromatin dysregulation in two stomach carcinoma patient samples.

Tumor heterogeneity remains a formidable challenge in cancer research and treatment, manifesting from inter-patient variability to intra-tumor clonal diversity. This heterogeneity complicates understanding of disease progression and therapeutic response. At the molecular level, chromatin dysregulation is a key driver of heterogeneity: aberrant chromatin states influence gene expression, cell identity, and tumor microenvironment (TME) interactions, all of which contribute to resistance pathways and metastasis.

Spatial ATAC-seq combines chromatin accessibility profiling with spatial resolution, allowing researchers to identify regulatory elements active in specific regions of a tissue. By coupling high-throughput sequencing with spatial barcoding approaches, the AtlasXomics platform enables:

• 10µm resolution: Identification of epigenetic states at the single-cell level.

• Spatial context: Mapping chromatin landscapes to morphological features.

• Comprehensive analysis: Integration of gene activity scores, motif enrichment, and peak accessibility with spatial representation.

These capabilities distinguish spatial ATAC-seq from bulk or single-cell ATAC-seq by preserving the physical location of cells, enabling deeper insights into tumor biology.

Mapping cells in TME

This dataset demonstrates how spatial ATAC-seq can disentangle cell state differences between tumor and surrounding normal tissue. By quantifying chromatin accessibility and inferring gene activity scores, it becomes possible to delineate distinct biological compartments and cell types, such as immune, muscle, and tumor-specific populations.

The UMAP projection below highlights major clusters including tumor, mucosa, muscle, and mixed tumor-immune states. When integrated with spatial mapping, these clusters align with morphological regions of the tissue, revealing how chromatin accessibility defines functional neighborhoods within the tumor microenvironment.

Such spatial alignment is essential for understanding how local chromatin states shape cellular interactions, influence immune infiltration, and drive tumor heterogeneity.

Deciphering Spatial Chromatin Dysregulation in TME

Chromatin accessibility reflects the interactions of DNA and transcription factors, and therefore a lens through which one can study underlying cell states in a tumor that lead to specific gene expression programs.

Increased accessibility may allow cancer cells to activate survival programs, upregulating anti-apoptotic genes (e.g., BCL2), stress response regulators, and genes promoting immune suppression (e.g., PD-L1, IDO1/IDO2 ) , enabling adaptation to therapy and ultimately, resistance. Resistant clones can exhibit unique chromatin landscapes with enriched accessibility at loci controlling drug efflux, transcription factors (e.g., MYC), and epigenetic modifiers.

Targeting chromatin regulators, such as histone deacetylases (HDACs) or bromodomain and extraterminal domain (BET) proteins, holds promise as a therapeutic strategy to counter these adaptations and re-sensitize tumors to treatment.

Differential Gene Accessibility Defines Distinct Tumor Compartments

Differential gene accessibility can reveal how distinct tumor compartments are functionally organized. By mapping the spatial distribution of highly differentially accessible genes, researchers can identify how cellular neighborhoods are poised to take on specialized roles of specific TME niches,such as immune hot vs. cold, invasive, or hypoxic environments.

Looking for Drivers of Disease

Examining motif enrichment in tumors compared to matching normal tissue reveals the regulatory elements driving cancer progression. Differential motif analysis reveals transcription factors and expression programs uniquely activated in tumors, shedding light on the epigenetic reprogramming that underpins oncogenesis. This approach can identify key cancer drivers, such as transcription factors supporting unchecked cell growth or immune evasion, and distinguish regulatory pathways associated with therapy resistance.

Additionally, comparing motif enrichment between tumor and normal tissues helps delineate how chromatin accessibility shifts contribute to tumor heterogeneity, highlighting the diverse adaptations of cancer cells to their microenvironment.

By pinpointing these unique regulatory elements, researchers can uncover novel biomarkers for early detection and identify region-specific therapeutic vulnerabilities.

Explore yourself

Spatial epigenomics presents a powerful framework for linking chromatin states to the physical architecture of tissues. By layering accessibility, histone modifications, and transcriptional readouts onto spatial coordinates, these datasets reveal how regulatory programs shape disease progression, tissue development, and therapeutic response.

• Visit atlasxomics.latch.bio/data-portal to visualize and explore yourself.

• Contact colinn@atlasxomics.com to learn how you can perform single-cell spatial CUT&TAG, or ATAC-seq in your experiments.

• If you are a kit or solution provider interested in setting up your own portal and bundle it with analysis packages, contact alfredo@latch.bio

• Release of a 25 million cell spatial atlas spanning all 11 major spatial vendors: including 10X Genomics Visium, Visium HD, Xenium, Bruker Spatial CosMx, STOmics Stereo-seq, Vizgen MERSCOPE, Takara Bio Seeker, AtlasXOmics DBiT-seq, Spatial Genomics GenePS (seqFISH), Element Biosciences AVITI24, and Singular Genomics G4X.

• A public data portal with annotated H5ADs ready to visualize and download

• A spatial vignette to allow researchers to continue with downstream steps, such as cell type annotation, ligand-receptor analyses, or multi-sample differential gene expression

• Tiled AnnData file format and visualization ecosystem to view large counts directly from a blob store without spinning up a server

Engineering biology is moving into an era of data driven and unbiased discovery. Large volumes of high quality measurements are used to tease out targets, disease mechanisms and molecular designs otherwise difficult to identify with the unaided eye.

Spatial assays provide rich information - precise molecular position and paired high resolution images - useful in modeling cell-cell interactions or emergent properties of entire tissues. These additional measurements will be necessary for AI labs to model phenomena beyond binding/structure or for data driven translational teams to mine new hypothesis space.

Scattered across the Internet are thousands of public spatial transcriptomics studies that research everything from 3D reconstruction of human embryos to tumor microenvironments for major cancer types. These datasets might help us develop a new class of therapeutics, but their potential remains locked away due to the expensive human labor required to clean and standardize them.

While there have been previous initiatives for large-scale curation, their efforts have been primarily focused on single-cell sequencing. This leaves statistical models blind to a critical dimension: spatial organization. By linking expression to physical location, we give models “eyes” to learn how cells interact with neighbors, how niches are structured, and how disease architectures emerge. This spatial context transforms fragmented molecular profiles into coherent maps of human biology, potentially opening avenues to therapeutic targets that dissociated single-cell data cannot reveal.

25M Spatial Cell Atlas for Human Anatomy

To tackle the curation challenge, the LatchBio engineering team aggregated 25M spatial transcriptomes from public human studies covering 45 tissue types, 63 diseases and 11 spatial technologies. To our knowledge, this is the largest open access sequencing based spatial atlas available to date.

The team used latch-curate, an human-in-the-loop curation framework that boosts annotation quality and consistency by incorporating information from entire papers and unstructured supplements.

A detailed whitepaper on latch-curate's design and capabilities is available at latch.bio/latch-curate.

Each dataset is published on console.latch.bio/data-portal as an annotated H5AD, structured with standardized ontologies for assay, cell type, condition, disease, tissue, patient ID (if available), and organism. Author annotations from the original paper are also included when available.

Spatial toolkit for scientists to go further

While curated datasets provide the substrate for scientific inquiry, more sophisticated downstream analyses are often required to extract knowledge and generate new hypotheses.

The Latch spatial toolkit provides scientists with step-by-step workflows to further analyze curated datasets or new studies from their own raw data.

The analysis can be broken down into 5 steps:

1. Counts construction from raw instrument data

2. Image alignment

3. Quality control, filtering, transformation

4. Multi-sample differential gene expression with pseudo-bulk analyses

5. Cell type annotation

6. Ligand-receptor analysis

Counts construction

Through partnerships with solution providers such as Takara Bio and AtlasXOmics, we build pipelines that convert raw instrument data into Scanpy-ready H5AD files. Alongside these, we generate PMtiles of morphology, H&E or IF stains, and transcript images. This collection can be loaded directly into the H5 viewer in Latch Plots for interactive exploration.

Image alignment

Spatial analysis requires precise segmentation of tissue regions. Once an H5AD is opened in the viewer, scientists can select images, define anchor points using the landmark system, and apply alignment. Both affine and STAlign modes are supported, with the latter designed for samples with severe tissue distortion.

Quality control, filtering, and transformation

Within the Plot notebook, scientists can run multi-sample QC, filter cells, detect highly variable and spatially variable genes, and apply Leiden clustering to define spatial domains of interest.

Multi-sample differential gene expression with pseudo-bulk analysis

After QC, scientists can launch pipelines such as cell2location for deconvolution in spot-based assays, RTCD, or marker-based approaches referencing PangloDB. These methods enable flexible annotation across technologies.

Ligand-receptor analysis

Once cell types are annotated, the next question is how they interact. Squidpy’s ligand-receptor module evaluates large numbers of spatial permutations, often at high computational cost. To increase focus and efficiency, scientists can use the viewer’s lasso tool to define regions of interest, ensuring interaction analysis captures only spatially proximal, biologically plausible events. Results are returned as plots of ligand–receptor pairs between cell types.

anntiles: tiled AnnData served directly from blob stores

To see a bird’s-eye view snapshot of the results generated by these tools, navigate to our Data Portal (console.latch.bio/data-portal) and click “Visualize” on our top featured datasets.

Our team has created a tiled AnnData file format (`anntiles`) and a suite of visualization tools that enable exploration of very large matrix-based counts data without spinning up a server. These tools will be released soon and are used to implement the notebooks exposed through the link above.

If you are a scientist interested in accessing these toolkits, please reach out and we will send you a link to access!

If you are a solution provider interested in white-labeling these tools and offering them to customers through a branded portal, contact alfredo@latch.bio. If you are interested in co-developing exciting new spatial tools and collaborating on our roadmap reach out to hannah@latch.bio.

What is a latch pod?

Latch pods are scientific computing environments built for biology. Pods let you run biological data science and machine learning applications in one place.

• A central system for JupyterLab, RStudio, and VSCode environments.

• A common set of tools for intuitive sharing of R and Python code.

• Collaborative notebooks for single-cell, spatial, image viewing, and more.

• Flexible control over CPUs, memory, and GPUs for different projects and users.

• Cloud billing features to monitor spending and set daily limits.

What is bring-your-own-compute?

Bring-your-own-compute is a new Latch architecture that allows you to:

• Connect your AWS environment

• Leverage your existing compute

• Specify geographic regions

• Use your cloud credits

All without data or compute leaving your servers.

How does it work?

High-level architecture

Bring-your-own-compute is part of a new latch architecture, which operates with a control plane and compute plane.

• The control plane includes services that Latch manages in your account.
  (e.g. Latch pods.)

• The compute plane is where your data is processed. There are two types of compute:

  • Latch Managed compute, which runs in your Latch account.

  • Bring-your-own compute, which runs in your AWS account.

Each Latch workspace has a storage bucket, which is in your AWS account.

Motivation and roadmap

We built this to deliver several benefits to our customers:

1. Ensure that data never leaves your cloud tenancy.

2. Allow you to use existing cloud provider credits.

3. Reduce network latency for users geographically far from us-west-2

4. Increase GPU availability by allowing compute hosts from other AWS regions

5. Improve cost efficiency by exposing a more diverse set of compute options

Based on interest, we’ll plan to continue supporting bring-your-own compute for the entire platform, including workflows, nextflow, and plots. Long-term, we’ll move towards support for GCP, Azure, and more compute providers.

How do I get started?

To get started, please comment or send a note to kyle@latch.bio and we’ll send you instructions. Thanks for reading.

This session features a live walkthrough of Plots—an early-access, browser-based tool for interactive spatial and single-cell analysis. Whether you're already using Seeker™ kits or exploring spatial biology for the first time, you'll get a practical look at how to unlock biological insights from complex datasets.

What you’ll learn:

• Remove off-tissue beads using background filtering for Seeker™ data

• Align and overlay H&E or immunofluorescence images onto spatial embeddings

• Explore and compare H5AD files across multiple samples

• Identify spatially variable genes and regions

• Detect differentially expressed genes across tissue zones

• Run ligand-receptor analysis in spatially defined domains

• Perform cell type annotation and deconvolution

• Map cellular trajectories in tissue context

Register now to see how these analyses work in practice.

Register for Webinar

Register for Webinar

What to Expect

The webinar will cover methods for:

1. Image Alignment for Seeker™ datasets with STAlign.

   Align serial-section H&E images to the spatial representation of the tissue.

2. Differential Gene Expression Across Spatial Regions

   Identify differentially expressed genes across Seeker™ or Trekker™ tiles.

3. Ligand-Receptor Analyses

   Detect communication between neighboring cell types using curated ligand-receptor pairs within selected regions.

4. Cell Type Annotation and Deconvolution

   Resolve cell-type composition from mixed transcriptomes using spatial-aware reference models.

…and more

Register to see a live demo of these analyses.

Register for Webinar

Register for Webinar

What You’ll Learn

Plots enables scientists to analyze complex spatial omics data through an intuitive, no-code interface. In this session, you’ll learn how to:

• Perform background removal

• Perform QC, filtering, and Leiden clustering

• Overlay H&E or immunofluorescence images on spatial embeddings

• Visualize and interact with H5AD files across multiple samples

• Use Lasso selection to define regions of interest

• Identify spatially variable genes

• Detect differentially expressed genes

• Conduct ligand-receptor analysis within selected spatial domains

• Perform cell type annotation and deconvolution

• Explore cellular trajectories in spatial context

Everything is accessible through a user-friendly graphical user interface. No programming required.

We look forward to seeing you there!

The past decade has witnessed many advancements in the scale and type of modality we can measure. Take the example of Takara Trekker, a clever technique to tag each nucleus within its native tissue environment with unique spatial barcodes using UV light. Or Parse Biosciences’ Penta kit, which uses a split-and-pool approach to barcode one to five million cells without the need to use microfluidics instruments for physical single-cell isolation. Even large-scale proteomics workflows, such as those from Olink or CITE-seq, now rely on antibodies conjugated to unique DNA sequences to convert protein abundance into a sequencing-compatible readout.

Whether you're measuring gene expression, chromatin state, protein abundance via DNA tags, or spatial context, the final step in these workflows is almost always the same: sequencing.

Originally built to read the human genome, modern sequencers have become general-purpose molecular readout devices—a universal endpoint for biology.

As of June 2025, the market features 37 sequencing instruments. To help navigate this landscape, we’ve mapped out the 10 key companies building these platforms, compared them across core technical attributes, and published a database to make the data easier to explore and debate.

You can access the database here, and if you see an error or want to contribute, leave a comment. This is meant to be a living and collaborative resource for the community.

In this article, we will explore:

• From Sanger to Today: The History of Sequencing Technologies

• Key Players in 2025: A Quick Rundown of Their Latest Chemistries

  • Oxford Nanopore: ONT’s Q30+ Kit14 (launched 2024)

  • Pacific Biosciences: HiFi Chemistry on Revio (launched 2023)

  • Pacific Biosciences: SPRQ Chemistry (launched late 2024)

  • Element Biosciences: Cloudbreak UltraQ (launched 2024)

  • Illumina: XBS-LEAP (launched 2023)

  • Complete Genomics: DNBSEQ with CoolMPS and HotMPS Chemistries (mature by 2025)

  • GeneMind: Lightning Chemistry (launched 2024)

  • Ultima Genomics: Solaris with ppmSeq (launched 2025)

• Comparison by Attribute

  • The U.S. seems to be trailing behind Chinese manufacturers in terms of throughput and speed.

  • Scientists need to navigate the cost per genome vs instrument price paradox.

  • Is Cost per GB all that matters? Beware of hidden infrastructure costs.

  • Multi-omics Support: Are sequencing machines evolving into something more?

We hope this will become a comprehensive resource for researchers as they navigate the NGS landscape in 2025.

From Sanger to Today: The History of Sequencing Technologies

1977 - 2005: The Foundation Years

The story of DNA sequencing begins in 1977 with Fred Sanger's chain-termination method, a breakthrough that first made reading the language of life possible. Sanger's initial triumph was modest by today's standards, decoding the 5,000-base genome of bacteriophage φX174, but it represented humanity's first glimpse into the digital code underlying biology. This pioneering method relied on clever chemistry: DNA polymerase would incorporate modified nucleotides that terminated chain extension, creating fragments of different lengths that revealed the sequence through gel electrophoresis.

For nearly three decades, Sanger's approach dominated the field through incremental improvements. The 1980s and 1990s brought automation via capillary sequencers, replacing labor-intensive gel systems with continuous processing machines. Yet even automated systems remained fundamentally limited; they were accurate but painfully slow, requiring years and tens of millions of dollars to sequence a single human genome. The Human Genome Project, completed in 2003, represented this era's culmination: a heroic international effort that demonstrated both genomics' transformative potential and the desperate need for faster, cheaper technologies.

2005 - 2010: The NGS Revolution

The mid-2000s brought the advent of “next-generation sequencing” (NGS), characterized by massively parallel short reads. 454 Life Sciences introduced a pyrosequencing platform in 2005 (read lengths ~100s of bases), and in 2006-2007 Solexa/ Illumina launched its sequencing-by-synthesis (SBS) platform, followed by Applied Biosystems’ SOLiD. These second-generation sequencers could generate gigabases of data in days, massively reducing the cost per base (dropping from ~$10,000 per megabase to mere cents). Illumina’s SBS technology quickly came to dominate the market, due to its accuracy and throughput. Over the past decade, Illumina has at times held ~80% of the sequencing market share.

2010s: The Rise of Long Reads (Third-Generation)

In the 2010s, third-generation sequencing technologies emerged, distinguished by the ability to sequence single molecules and produce much longer reads (thousands to tens of thousands of bases).

Pacific Biosciences pioneered this transition with their Single Molecule Real-Time (SMRT) sequencing platform in 2011. Their zero-mode waveguides created optical chambers where individual DNA polymerases could be observed incorporating fluorescent nucleotides. Oxford Nanopore Technologies took a different approach, threading DNA through protein nanopores and measuring electrical signal changes as bases passed through. When ONT's MinION launched mid-decade, it brought sequencing capabilities to a USB-stick-sized device.

These long-read platforms initially faced skepticism due to higher error rates, but the errors were random rather than systematic, making them correctable through coverage or consensus approaches. More importantly, long reads solved problems short reads couldn't address: de novo genome assembly of complex regions, large structural variant detection, and full-length isoform sequencing. As accuracy improved (PacBio's HiFi reads achieving over 99.9% accuracy, ONT's newer chemistries reaching ~99% single-read accuracy), long reads transitioned from specialized tools to mainstream platforms.

By the late 2010s, the cost of sequencing a human genome had fallen below $1,000, a milestone achieved well ahead of the trajectory of Moore’s Law.

2020s: Toward Multi-omic, Spatial, and Ultra-High-Throughput Sequencing.

We are now in a new phase defined by multi-omic compatibility, spatially-resolved sequencing, and ultra-high throughput machines. Sequencing is being integrated with other modalities: for example, Illumina and Singular Genomics are developing workflows to sequence not just genomes, but also spatial transcriptomics. Illumina also recently acquired SomaLogic, signalling its movement into the proteomics field.

Throughput has also exploded. Traditional “high-throughput” sequencers output on the order of a few terabases per run, but new instruments are extending this by an order of magnitude. Illumina’s latest NovaSeq X series, for instance, can output up to 16 terabases of data in a single run (26 billion reads per flow cell), and emerging players are pushing costs down further (more on this below). There is also a convergence of technologies: short-read companies are adding long-read or synthetic-long-read capabilities, long-read companies have launched short-read platforms, and entirely new chemistries are being introduced. All of this means researchers in mid-2025 face a rich but complex menu of sequencing options.

Key Players in 2025: A Quick Rundown of Their Latest Chemistries

Oxford Nanopore: ONT’s Q30 Duplex Kit14 (launched 2024)

Oxford Nanopore Technologies (ONT) sequences DNA using biological nanopores: tiny protein channels embedded in a membrane that detect DNA molecules as they pass through. As a single-stranded DNA molecule moves through the pore, it causes disruptions in ionic current. These changes are measured and interpreted by machine learning models to identify the underlying nucleotide sequence in real time.

Traditionally, ONT relied on simplex sequencing, where only one strand of a DNA molecule is read. While this allows for extremely long reads, often tens of kilobases or more, simplex reads suffer from several challenges. The signal measured by the pore reflects a short k-mer (typically 5 or 6 bases), making it difficult to resolve individual bases when homopolymers or repetitive regions are present. Because only one strand is read, there is no independent validation of each base, so errors-especially insertions and deletions-tend to persist.

As a result, simplex reads historically achieved modal accuracies of only Q15 to Q18, or about 97 to 98 percent accuracy.

The introduction of Q20+ Kit14 brought ONT’s first broadly available duplex sequencing chemistry. In duplex mode, both strands of a double-stranded DNA molecule are sequenced in succession using a specially designed hairpin adapter. The basecaller then aligns the two reads and compares them to correct random errors and resolve ambiguous regions. Because the errors in each strand are largely uncorrelated, this reconciliation process significantly improves confidence in each base. Simplex reads now achieve around Q20 (~99%) accuracy, while duplex reads regularly exceed Q30 (>99.9%), rivaling the performance of short-read platforms in base-level precision.

Practically speaking, this means that ONT with Q20+ Kit14 can now be used for workflows previously out of reach for nanopore due to accuracy concerns, such as low-frequency variant detection, cfDNA sequencing, or methylation-aware diagnostics, while still enabling applications Illumina cannot touch, like ultra-long read assemblies, haplotype phasing, or native DNA/RNA modification analysis.

Pacific Biosciences: HiFi Chemistry on Revio (launched 2023)

Pacific Biosciences (PacBio) developed a sequencing method called single-molecule real-time (SMRT) sequencing, which watches DNA being copied in real time, one molecule at a time. In this system, a single DNA polymerase enzyme is anchored inside a tiny well called a zero-mode waveguide (ZMW). These wells are designed so that only the very bottom, where the enzyme is working, is illuminated. As the polymerase adds each nucleotide, which carries a unique fluorescent tag, the system detects a quick flash of light and records which base was incorporated.

This technology allows PacBio to read very long stretches of DNA without needing to break them up into short fragments, but early versions of SMRT sequencing had a major drawback of low accuracy.

Because each molecule was only read once, errors from things like fluorescence misreads, dye noise, or enzyme slippage made the data noisy. Single-pass reads often had error rates around 10-15%, especially in homopolymer regions or sequences with complex structure.

To solve the accuracy problem in its original SMRT sequencing, PacBio developed HiFi reads (short for High-Fidelity reads), which combine the length advantages of long-read sequencing with the accuracy of short-read platforms.

Instead of reading a DNA molecule once, PacBio circularizes each DNA fragment using special adapters, forming a loop called a SMRTbell template. The DNA polymerase then continuously reads around the circular molecule, passing over the same insert multiple times, usually 10 to 20 passes. Each pass produces a slightly different version of the sequence due to random errors. By averaging these multiple observations, PacBio’s algorithm generates a circular consensus sequence (CCS) that is highly accurate.

This technique drastically improves read quality. HiFi reads typically range from 10 to 25 kilobases in length, and because they’re built from multiple subreads, they reach Q30 to Q40 accuracy (99.9-99.99%).

Starting at a modest $600M in revenue in 2023, long-read sequencing is projected to reach $1.34B in 2026, likely due to increased market trust and accuracy.

Pacific Biosciences: SPRQ Chemistry (launched late 2024)

Building on the success of HiFi sequencing, PacBio took a bet on their first multi-omics chemistry: SPRQ (pronounced “spark”), designed to extract both DNA sequence and regulatory information from the same molecule.

Unlike bisulfite or enzymatic methods that chemically modify or fragment DNA, SPRQ preserves long, native DNA molecules. A transposase enzyme is used to preferentially insert special adapters into open (accessible) regions of the genome. These adapters include 6-methyladenine (6mA) marks. Open chromatin tends to be more accessible to the enzyme, so this step selectively “labels” active regulatory regions.

As with standard HiFi sequencing, these DNA fragments are circularized into SMRTbell templates, which allows the polymerase to read the same molecule multiple times for high-fidelity consensus sequencing. PacBio’s polymerase begins copying the circular DNA in real time.

As it passes through modified regions (like 6mA or naturally occurring 5mC), the speed of base incorporation changes slightly. These kinetic "pauses" or slowdowns are subtle but measurable. The sequencing software detects these polymerase slowdowns and matches them to specific base positions. It then combines: base sequence from HiFi reads (Q30-Q40), 6mA tagging patterns (open chromatin), and methylation signals like 5mC (native epigenetics). In just one sequencing run, SPQR provides high-accuracy DNA sequence, maps of chromatin accessibility, and detection of base modifications like 5mC and 6mA.

Element Biosciences: Cloudbreak UltraQ (launched 2024)

Element’s AVITI system uses sequencing-by-synthesis (SBS), but with a unique chemistry built around rolling circle amplification (RCA). Instead of reading a single DNA strand, AVITI circularizes DNA fragments and amplifies them into dense clusters called polonies, thousands of identical copies of the same DNA molecule in one spot. As the sequencer adds fluorescently labeled bases one at a time, it captures images and averages the signal across all the identical copies in the cluster. This built-in redundancy helps reduce random noise and improves overall base-calling accuracy.

In 2024, Element introduced Cloudbreak UltraQ, a high-accuracy upgrade to its standard chemistry. UltraQ builds on the same core principles but increases the number of observations per base and improves the enzymes and dyes used during synthesis. This results in sharper signals and cleaner base incorporation, which the AVITI’s software can use for even more precise error correction. With UltraQ, the system reaches Q50+ accuracy (just one error in 100,000 bases!)

Illumina: XBS-LEAP (launched 2023)

Illumina’s newest sequencing chemistry (as of 2023-2025) is XLEAP-SBS™, which was

introduced with the NovaSeq X series instruments. XLEAP-SBS (previously code-named “Chemistry X”) is an overhauled SBS chemistry engineered for faster cycle times and higher accuracy. Not a lot of technical details are available online on how the chemistry actually works, but Illumina claims the following improvements: a new thermostable polymerase, optimized fluorescent terminator nucleotides, and upgraded clustering chemistry on patterned flow cells . The result is roughly 2× faster incorporation speed per cycle and up to 3× greater basecalling accuracy. In fact, NovaSeq X flow cells can generate ≥85% of bases at Q40 or higher (99.99% accuracy), in comparison to the Q30 standard of prior sequencers. By

late 2023 Illumina enabled 2×300 bp reads on NovaSeq X flow cells (primarily for microbiome and metagenomics users needing longer coverage). Additionally, Illumina has back-ported XLEAP-SBS to mid-range instruments. An early 2024 update allows NextSeq 1000/2000 systems to use XLEAP-SBS with new P4 flow cells (~500 Gb output), indicating Illumina’s intent to make the chemistry standard across its lineup.

Complete Genomics: DNBSEQ with CoolMPS and HotMPS Chemistries (mature by 2025)

Complete Genomics’s sequencing platforms are built around a core technology called DNA Nanoball sequencing (DNBSEQ). Unlike other short-read platforms that use bridge amplification, MGI circularizes DNA fragments and amplifies them using rolling circle replication. This produces compact DNA nanoballs, each consisting of hundreds of identical copies of a single DNA molecule. These nanoballs are then loaded onto a patterned flow cell at high density, with each occupying a precise, separate position. Because each nanoball contains many copies of the same fragment, signal averaging becomes highly effective, allowing for accurate base detection with minimal cross-talk between spots.

Traditionally, MGI sequencers used a sequencing-by-synthesis (SBS) approach similar to Illumina’s, with fluorescently labeled nucleotides added one base at a time and imaged to determine the sequence. However, MGI has developed two distinct chemistries to further improve speed, accuracy, and data quality: HotMPS and CoolMPS.

HotMPS is an updated version of SBS that uses directly labeled nucleotides with enhanced polymerase enzymes. The chemistry has been optimized to reduce incorporation bias and fluorescence noise, delivering faster cycle times and improved base-calling accuracy. It's the default chemistry for high-throughput runs on platforms like the T7 and T20, where large-scale projects require both speed and scalability.

CoolMPS, by contrast, takes a different approach. Rather than using fluorescently labeled nucleotides, CoolMPS incorporates unlabeled bases during sequencing. After each incorporation step, fluorescently labeled antibodies that specifically recognize the incorporated base are added to detect which base was added. This indirect detection strategy reduces chemical "scarring" and dye-related artifacts, resulting in cleaner signal and longer read lengths with fewer errors. Because the nucleotides themselves are unmodified, polymerase activity is more consistent and incorporation fidelity is higher.

Together, these chemistries support read lengths up to 2×300 bp, with over 85% of bases typically achieving Q30 quality. On the highest-end system, the DNBSEQ-T20×2, MGI platforms can output up to 72 terabases in a single 80-hour run, currently the most in the industry! This scalability, paired with high accuracy and unique chemistry options, makes MGI a strong contender in applications like population-scale genomics, clinical diagnostics, and spatial transcriptomics.

GeneMind: Lightning Chemistry (launched 2024)

GeneMind’s SURFSeq Q uses classic sequencing-by-synthesis (SBS) on patterned flow cells, where DNA fragments amplify into dense, equally spaced clusters across the chip. Each cycle, fluorescently labeled nucleotides are added to the clusters, imaged, and then cleaved to read the next base, the same core process as Illumina-style sequencers.

To significantly boost both speed and quality, GeneMind introduced Lightning Chemistry in 2024. This chemistry improves three core areas: signal clarity, cycle speed, and base-calling accuracy. Brighter fluorescent labels and refined polymerase enzymes allow faster and cleaner nucleotide incorporations. The combination enables ultra-high throughput. SURFSeq Q delivers up to 14 Tb per run (≈9 Tb/day), while maintaining exceptional accuracy, with 90% of bases at Q40 or higher.

Ultima Genomics: Solaris with ppmSeq (launched 2025)

Ultima Genomics takes a unique approach to short-read sequencing by replacing traditional flow cells with a spinning silicon wafer etched with billions of tiny nanowells. Each well is designed to hold a single DNA fragment, which is anchored using special adapters during library preparation. Once loaded, the wafer spins rapidly to distribute sequencing reagents evenly across its surface, a process that also minimizes fluidic complexity and speeds up reaction cycles.

Ultima’s sequencing chemistry follows a method called sequencing-by-addition. Instead of flowing in all four nucleotides (A, T, C, G) at once (as is common in sequencing-by-synthesis platforms) Ultima flows only one base at a time. For example, in the first cycle, only adenine (A) nucleotides are introduced. If the complementary base on a DNA fragment is thymine (T), the polymerase incorporates an A and emits a fluorescent signal. If not, no incorporation occurs and no signal is recorded. This single-base flow cycle is repeated sequentially (A, then C, then G, then T) across hundreds of cycles to reconstruct the DNA sequence one base at a time.

To enhance basecalling accuracy, especially in error-prone regions like homopolymers, Ultima employs a proprietary method called ppmSeq (plus/minus strand sequencing). This technique sequences both strands of the same DNA molecule independently. The forward and reverse reads are then compared, allowing the system to detect and correct mismatches or insertion-deletion (indel) errors. By cross-validating the two strands, Ultima improves variant calling accuracy and read confidence without sacrificing speed or cost.

Thanks to its simplified chemistry and real-time error correction, Ultima achieves variant-calling performance comparable to short-read leaders like Illumina, with base-level Phred quality scores approaching Q30-Q40, or 99.9% to 99.99% accuracy.

Singular Genomics

Singular Genomics’ G4X platform introduces a hybrid of high-throughput sequencing and spatial biology in a single system. At its core, G4X uses sequencing-by-synthesis (SBS), where fluorescently labeled nucleotides are incorporated one base at a time and recorded by an optical system. But what sets G4X apart is its ability to perform in situ sequencing directly on preserved tissue samples, using a method called Direct-Seq.

With Direct-Seq, FFPE (formalin-fixed paraffin-embedded) tissue sections can be sequenced without the need to extract and convert RNA into cDNA. Instead, the nucleic acids remain physically anchored in the tissue, and sequencing happens right where the molecules naturally reside. This is achieved by flowing specially designed polymerases and fluorescent nucleotides over the sample. As the enzyme incorporates each nucleotide, it emits a flash of light that is recorded by the instrument’s camera, identifying the added base and its precise location within the tissue.

This direct sequencing preserves spatial context while revealing gene expression, allowing researchers to detect RNA transcripts without disrupting the surrounding tissue architecture. In parallel, the G4X system can also perform protein detection using fluorescently tagged antibodies and identify genomic variants, such as single nucleotide changes or small insertions and deletions, from the same section.

By layering RNA, protein, and mutation data into one integrated dataset, G4X is the first to enable full spatial multi-omic profiling in the same sequencer.

Short-read platforms are adding long-read capabilities.

Historically, short-read sequencers like those from Illumina dominated the market with high accuracy and massive throughput, but they struggled to resolve complex regions of the genome due to limited read lengths. To overcome this, Illumina introduced Complete Long Reads (CLR), a synthetic long-read approach that uses molecular barcodes to tag long DNA fragments, which are then sequenced as short 2×150 bp reads and computationally stitched back together.

This enables users of NovaSeq X and 6000 to generate long-read-like data without investing in new hardware. The reconstructed long reads achieve Q30-Q40+ accuracy (99.9-99.99%).

Other short-read companies are following a similar strategy. Element Biosciences, for instance, acquired Loop Genomics to offer barcode-based long-read reconstruction on the AVITI platform. The platform’s native short-read chemistry already reaches Q40 accuracy (99.99%), making it well-suited for high-precision long-read assembly.

These developments reflect an interesting trend. Short-read platforms are expanding into long-read territory by offloading complexity to molecular barcoding and software, while preserving the accuracy and cost-efficiency advantages of short-read sequencing.

Long-read platforms are also doing short-reads.

Oxford Nanopore Technologies (ONT) introduced Short Fragment Mode to support DNA fragments as short as 200-500 bp, enabling workflows with degraded or low-input samples like FFPE tissue or microbial DNA. Combined with Duplex Sequencing, which boosts accuracy by sequencing both strands of a molecule, ONT now supports applications once reserved for short-read platforms.

PacBio launched Onso, a dedicated short-read sequencer using a novel sequencing-by-binding chemistry. Instead of incorporating nucleotides, this approach detects fluorescently labeled bases as they bind reversibly to the template strand, reducing signal noise and improving accuracy. Onso achieves Q40+ accuracy and is aimed at precision-focused markets like oncology and NIPT.

Though these are technically impressive developments, it’s still unclear whether long-read companies can compete head-on in the short-read market. ONT’s and PacBio’s efforts appear more like adaptations to support existing users, those who already own a MinION or Revio, and want to handle occasional short-read workloads, rather than serious attempts to replace Illumina or AVITI for core short-read use cases.

Comparison by Attribute

Disclaimer: This section is built on our CSV, which aggregates public and online information. While some providers have transparent spec sheets on their websites, others don’t and require direct sales contact. In such cases, we tried our best to provide estimates. Please reach out to hannah@latch.bio to propose any corrections.

The U.S. seems to be trailing behind Chinese manufacturers in terms of throughput and speed.

One headline metric is maximum data output (throughput). This determines how many samples or how deep a sequencing run can go in a given time.

The current champions in output are Complete Genomics’s DNBSEQ-T20 (or T10 or T7), Illumina’s NovaSeq X, GeneMind’s SURFSeq Q, ONT’s PromethION 48, and Ultima’s UG 100 - all capable of multi-terabase outputs per run:

• DNBSEQ-T Series: Topping the chart is Complete Genomics, a subsidiary of Complete Genomics in China, which dominates when it comes to throughput per day. Their DNBSEQ-T20 instrument yields 72 Tb data for one 80-hour run, translating to ~21.6 Tb/day. The T10 variant achieves even higher total output at 76.8 Tb per run over 106 hours (~17.4 Tb/day), while the T7 offers 7 Tb in 24 hours (7 Tb/day).

• SURFSeq Q: GeneMind's high-end platform achieves an impressive 14 Tb total output in just 36 hours, translating to over 9.3 Tb/day. This positions it as the second-fastest daily throughput platform, though with lower total capacity than the largest MGI systems. A smaller SURFSeq configuration matches the DNBSEQ-T7 at 7 Tb/day.

• NovaSeq X Series: Illumina's flagship production platform delivers up to 16 Tb per run in 48 hours, achieving 8 Tb/day at peak configuration. Multiple flow cell options provide flexibility, with smaller configurations generating 3-8 Tb per run. The platform balances throughput with established ecosystem support, though it trails MGI's peak daily rates.

• UG 100: Ultima Genomics' cost-focused platform generates 3.6 Tb per run in 12-14 hours, achieving approximately 6.2 Tb/day. While lower in absolute output, this platform targets the $100 genome milestone through exceptional cost efficiency rather than maximum throughput.

• PromethION Series: Oxford Nanopore's long-read flagship achieves 7 Tb total output but over 72-hour runs, resulting in 2.3 Tb/day for the 24-flow cell system. The 2-flow cell variant produces 580 GB over similar timeframes. While lower in raw throughput, these platforms enable applications impossible with short-read systems through ultra-long reads exceeding up to 4 MB.

In the upper quadrant of the market, we have Ultima Genomics’ UG100 ($1.5M), Illumina’s NovaSeq X ($985,000) and NovaSeq X Plus ($1.25M), with NovaSeq 6000 having a similar price tag as NovaSeq X, but with a different range of throughput. Complete Genomics’s DNBSEQ-T7 especially stands out, since it delivers 7 Tb/day—nearly matching NovaSeq X Plus’s 8 Tb—but costs only ~$600K, half the price. Even in the mid-throughput range, DNBSEQ-T1+ comes at $299,000, almost a third of the cost of a NovaSeq 6000. (Note: Complete Genomics doesn’t publish its costs online, but these prices have been confirmed by a representative from Complete Genomics)

The market is heavily clustered in the low-to-mid throughput and cost range. Let’s zoom into them.

In the low-throughput range, most instruments output under 500 Gb/day, with a wide spread in pricing. Illumina offers the broadest lineup—from the low-cost iSeq to the mid-tier NextSeq 1000—ranging from ~$20K to over $300K. Oxford Nanopore and Thermo Fisher offer alternatives like MinION, GridION, and Genexus, with varying form factors and costs around $100K–$300K.

Element Biosciences and Complete Genomics enter at slightly higher throughput levels (100–400 Gb/day) with machines priced near $300K–$400K, offering better output per dollar. PacBio’s Revio is the clear outlier (producing ~400 Gb/day at ~$600K) but includes HiFi accuracy and native methylation calling, justifying its premium.

Scientists need to navigate the cost per genome vs instrument price paradox.

The most expensive machines often deliver the lowest cost per genome, while affordable instruments can carry surprisingly high per-sample costs. You can achieve low cost per gigabase, but only with instrument investments in the millions of dollars range. Looking at cost per genome alone can be misleading.

Below, we compare cost per genome across various providers on the market. To calculate cost per genome, we:

• Gather maximum data yield per run from instrument provider websites

• Obtain maximum reagent cost per run from provider websites or news report estimates when unavailable

• Calculate cost per gigabase = maximum reagent cost ÷ maximum data yield

• Assume 30X genome coverage requiring approximately 100 Gb of sequencing

• Calculate cost per genome = cost per gigabase × 100

In the realm of short-read sequencing, leading this race is Ultima's $1.5M UG 100, which achieves the lowest cost at $80 per genome, while Illumina's $1.25M NovaSeq X delivers $103-268 per genome depending on configuration.

Complete Genomics doesn’t sell their DNBSEQ-T20 directly. Instead, pricing is amortized over reagent contracts (e.g., $0.99/Gb for 72 Tb/run or $99/per genome), good for mega-scale centers sequencing >50k genomes/year.

In the middle tier, platforms like the NovaSeq X 1.5B, DNBSEQ-T7, and AVITI strike a balance between upfront investment and per-genome cost. These instruments typically fall in the $300K to $1M range and bring sequencing costs down to around $150-300 per genome. They’re a good fit for labs that don’t need ultra-high throughput but still want to process thousands of samples a year efficiently. Many offer flexible flow cell configurations, so you’re not locked into massive runs to make the economics work.

Element Biosciences' AVITI emerges as a fierce contender in the mid-range market. At $289,000, the instrument delivers $200 per genome costs that directly compete with Illumina's NovaSeq X 25B ($206 per genome) while requiring less than one-third the capital investment ($985K for NovaSeq X). Even AVITI's lower-throughput configurations maintain competitive economics, with cost per gigabase ranging from $5.60-$7.20, closely matching the NovaSeq 6000's $4.55-$6.23 range across S1, S2, and SP flow cells. This positioning allows institutions to access near-production-scale economics without the substantial upfront investment typically required for high-throughput sequencing.

While short-read platforms continue to drive costs down across a range of throughput tiers, long-read technologies are following close behind, offering increasingly competitive pricing with the benefit of spanning complex regions.

Long-read whole-genome sequencing costs are difficult to pin down because they hinge on several interacting factors. While Oxford Nanopore advertises roughly $345 for about 100 Gb and PacBio quotes around $500 for ~30× human genome coverage on its Revio system, actual price depends on flow-cell yield (Nanopore flow cell output ranges from ~100 Gb to 290 Gb), the accuracy mode chosen (cheaper Q20 simplex reads versus higher-quality Q30+ duplex reads), and application-specific depth or accuracy requirements.

Is Cost per GB all that matters? Beware of hidden infrastructure costs.

Converting electrical signals and optical data into usable format is no easy task. While cost per gigabase has become the dominant metric for comparing sequencing platforms, this narrow focus obscures the substantial computational infrastructure investments required to transform raw sequencing data into useful insights.

Once bases are called, labs must invest in compute infrastructure, software licenses, data transfer pipelines, storage, and ongoing maintenance to extract value from the output.

What do these hidden costs include?

1. Primary analysis: basecalling, demultiplexing

2. Secondary analysis: alignment, variant calling, assembly

3. Data storage: terabytes of raw + processed data per run

4. Compute infrastructure: on-prem clusters or cloud credits

5. Software licenses: proprietary tools (e.g. DRAGEN)

6. Data transfer: ingress/egress fees for cloud platforms

7. Compliance: secure, reproducible pipelines for CLIA/CAP/ISO especially for diagnostics use cases

At AGBT 2025, Illumina CEO Jacob Thaysen captured this shift:

"Customers are not only buying a product anymore…they're buying into a whole ecosystem… applications, workflows, informatics… That is the game."

Below, we compare how major sequencing vendors address these hidden infrastructure costs through engineering design and software strategy.

DRAGEN: FPGA-accelerated genomics

Illumina recognized the challenges in storage and processing of genomic data early, acquiring Edico Genome and the DRAGEN Bio-IT platform in 2018 to architect a solution.

DRAGEN uses highly reconfigurable field-programmable gate arrays (FPGA) to provide hardware-accelerated implementations of genome analysis algorithms, reducing processing time for a 40× whole human genome from over 8 hours with traditional open-source solutions to approximately 34 minutes.

The engineering behind DRAGEN's performance is also quite interesting. The NovaSeq X Plus features dual socket AMD EPYC 7552 processors with 96 x86_64 cores, four Xilinx Alveo U250 FPGA cards, and 1.5 TB of RAM, enabling on-instrument DRAGEN processing to output fully processed 30× WGS samples approximately once every four minutes while drawing power from a single standard 200-240 volt AC plug (about the same energy required to propel a Tesla Model 3 half a kilometer). The resulting compute cost averages <10 US cents.

However, DRAGEN's impressive technical capabilities come with significant licensing costs. Since Illumina doesn't benefit from FPGA hardware sales, they've adopted an aggressive software licensing model. When running DRAGEN on cloud platforms like AWS, the license cost represents approximately 80% of the total computational expense, a pricing structure that can make cloud-based processing prohibitively expensive for large-scale projects.

Oxford Nanopore and PacBio Revio: Onboard GPUs right inside the device

Oxford Nanopore takes a different approach by building GPUs directly into its high-throughput sequencers, so there’s no need for separate compute clusters or extra software to process the data.

The PromethION P48 now comes with four NVIDIA A100 GPUs, which speed up basecalling by more than 50% compared to older models. This means the sequencer outputs already basecalled data. You get usable FASTQ files right off the machine, ready for analysis. Each run can generate up to 10-13 terabases of data over 72 hours. And because the data is processed in real time on the device, there's no need to move massive raw signal files (which can be 2-3X larger) to the cloud or external servers.

Similarly, PacBio’s Revio also includes onboard NVIDIA A100 GPUs. The system handles basecalling, error correction, and consensus (HiFi) reads on-device, delivering BAM/FASTQ files ready for analysis.

Each SMRT Cell produces ~120 gigabases in ~24 hours. With four cells in parallel, that's up to 480 Gb/day, enough for 1,300 human genomes per year at 30× coverage, all for under $500 per HiFi genome using their SPRQ chemistry and reagent pricing.

Element cloud-first strategy

Element's AVITI platform integrates natively with Amazon Omics, streaming data directly to customer S3 buckets during sequencing and enabling immediate parallel analysis without costly subscription services or intermediate data transfers. Bases2Fastq Ready2Run workflow on Amazon Omics costs approximately $3 to generate FASTQ files for a 300‑G base run (2×150 bp) and completes in about 1 hour for demultiplexing, with the full pipeline (FASTQ → BAM → VCF) finishing in ~2 hours.

Ultima’s hybrid approach

Each UG 100 sequencer comes with built-in high-performance compute that runs standard pipelines like basecalling, alignment, and variant calling directly on the instrument. For downstream analysis, users can upload the output to the cloud and run Ready2Run pipelines on AWS HealthOmics, including DeepVariant ($2 per genome in US regions), Sentieon DNAscope, and GATK workflows. For advanced users, Ultima offers open-source Terraform and WDL modules that make it easy to deploy private workflows using their own AWS storage, compute, and custom references.

Software competitiveness will become more important in the future.

As companies integrate with more multi-omics modalities and become more verticalized, building an ecosystem of robust bioinformatics tools and doing so fast will be important. Prime example is Illumina who’s leading this trend by releasing the Integrated Connected Multiomics platform.

The dynamics in the sequencing market mirror the early smartphone era: As genome sequencing costs plummet, the device matters less than the ecosystem of multiomics kits and tools around it.

Platforms that simplify the transition from raw data to insight will ultimately win the loyalty of users long-term.

Multi-omics Support: Are sequencing machines evolving into something more?

What began as instruments designed to read DNA has quietly transformed into something far more ambitious: platforms that can simultaneously peek into multiple layers of cellular life. Instrument providers seem to be asking a new question: not only "how fast can we sequence DNA?" but also "how completely can we understand a cell?"

This paradigm shift became clear in February 2025 when Illumina announced their spatial technology program. Rather than just reading sequences, their platform maps entire transcriptomes while preserving the spatial context, profiling millions of cells on two slides, supported by Connected Multiomics software designed for researchers without bioinformatics expertise.

Singular had already signaled this direction a year earlier with their G4X announcement in February 2024. Their instrument handles both traditional sequencing and spatial analysis seamlessly. Through Direct-Seq technology, they read RNA directly within intact FFPE tissue without library preparation, simultaneously detecting transcripts, imaging proteins, and reading mutations while maintaining spatial context. Early users have demonstrated 3D tissue reconstructions analyzing over 6 million cells from a single flow cell.

Element also pushed this concept with their AVITI24, performing what they call "5D multiomics": simultaneous detection of DNA, RNA, proteins, phosphorylated proteins, and cell morphology from the same sample in under 24 hours. Their upcoming Direct In Sample Sequencing eliminates library prep entirely by targeting RNA molecules directly within cells. Teton assays focus on specific pathways, currently targeting 350 genes and 50 proteins in MAPK signaling.

PacBio completed this picture with their SPRQ chemistry, launched late 2024. Using 6mA markers in Fiber-seq, every run becomes multiomics by default, simultaneously calling DNA sequence, methylation, and chromatin accessibility. Researchers can extract four molecular views (genome, methylome, epigenome, and transcriptome) from a single sample.

Instrument makers are racing to integrate upstream prep and downstream analysis directly into their platforms through aggressive acquisitions and internal development. Illumina acquired Fluent BioSciences mid 2024 to bring single-cell library prep in-house, then followed with a $425M acquisition of SomaLogic mid 2025 to add proteomics assays to its sequencing ecosystem. Thermo Fisher made a similar move by acquiring Olink for $3.1B late 2023, expanding into high-throughput protein biomarker analysis. Meanwhile, Complete Genomics has internally developed a full-stack system that includes its DNBSEQ sequencing platforms, MGIFLP library prep systems, and a growing multi-omics portfolio spanning spatial transcriptomics, methylation, whole-genome sequencing, single-cell, exome, and proteomics applications.

Starting as DNA readers, these instruments increasingly look more like a “cellular omniscope”, aiming to capture multiple modalities of a cell’s biology.

Read Length

Different applications demand different read lengths. The table below summarizes recommended configurations and throughput per sample for common short-read workflows:

Most short-read sequencing platforms today use a 2x150bp format, which means they sequence 150 bases from each end of a DNA fragment. It also aligns well with common DNA fragment sizes used in WGS libraries, which typically range from 350 to 400 bp. Leading high-throughput instruments, including Illumina's NovaSeq X, Complete Genomics' DNBSEQ-T7/T1+, and GeneMind's SURFSeq Q, all use this 2×150bp configuration.

Mid-throughput instruments like the NextSeq 2000/1000 and MiSeq i100 can support 2×300bp reads, but this extended read length comes at the cost of reduced throughput. This trade-off stems from fundamental limitations in Illumina's sequencing-by-synthesis (SBS) chemistry and bridge amplification process.

Illumina's bridge amplification requires both ends of a DNA fragment to attach to the flow cell surface, forming a bridge structure. The DNA then undergoes in-place amplification to create clusters of identical fragments. This process works optimally with shorter DNA fragments because longer fragments struggle to achieve the necessary flexibility for both ends to efficiently bend and hybridize to the surface. When bridging fails, fragments cannot amplify into clusters, resulting in reduced cluster density, weaker signal intensity, and lower overall yield.

On the contrary, new players in the field such as Roche are breaking the trend with their new SBX system that’s in R&D. Roche promised in their AGBT presentation with “simplex reads ranging from 50 bp to >1,000 bp”, “duplex reads (~150–350 bp inserts) delivering intramolecular consensus with Q39 accuracy”, and even “midi reads reaching 1,200-1,500 bp for longer contiguous coverage. SBX also claims it can sequence 7 human genomes (30× depth) in just one hour, over 5 billion duplex reads per hour. These newer players are likely putting competitive market pressure on Illumina as they are likely bringing the best from both worlds: supporting longer fragments, and while also enabling high-throughput sequencing.

Parting Thoughts

The sequencing landscape in 2025 reflects several converging trends.

Sequencing providers are rapidly verticalizing, transforming from instrument vendors into “full-stack” companies that control everything from sample prep, assays, to data interpretation and insights.

The traditional boundaries between short and long reads continue to blur. Short-read platforms now generate synthetic long reads through molecular barcoding, while long-read companies launch dedicated short-read instruments. Mid-length chemistries promise to bridge the gap with kilobase reads at short-read accuracy levels.

The relentless drive toward lower cost per gigabase shows no signs of slowing. The $100 genome milestone is within reach, with Ultima targeting even lower thresholds potentially approaching $50 per genome by the late 2020s. As reagent costs become commoditized, ecosystem support may emerge as the primary differentiator over raw sequencing economics.

Long-read technologies are poised for clinical adoption as accuracy improvements and cost reductions to <$1000 per genome make long-range genomic profiling possible. The vision of single-platform workflows capturing SNPs, structural variants, methylation, and RNA expression simultaneously is materializing through providers' push for better enhanced chemistry and error correction algorithms.

Finally, sequencers are becoming multiomics platforms. These instruments now capture DNA sequence alongside spatial context, protein abundance, and epigenetic modifications in single runs.

All roads still lead to sequencing machines. But these machines are becoming far more powerful than their original architects ever imagined.

• A 30M cell observational atlas spanning 150 diseases, 200 tissues and 27 technologies

• A partnership with expert data curators Pythia Biosciences and Miraomics

• An agentic human-in-the-loop Python framework for mass scRNA-seq curation

We’re approaching a new era of biology too complex to navigate with unaided human cognition. The industry needs large volumes of structured molecular data to develop a new class of foundation models and build large atlases. It is becoming clear that data is the oil of modern biotech.

Latch is addressing this need by building curation infrastructure, tools and delivery portals for data solution providers. By equipping our partners with powerful new tools, our ambition is to organize the public molecular information scattered across the Internet and provide it for immediate download on a usage basis.

A 30 Million Single Cell Atlas

The need for large scale data has been met initially by purpose-built industrial data-generation projects. These efforts are incredible but do not sample sufficiently broad observational space, especially for rare indications.

Public datasets remain the largest and most diverse reservoir of diseases, tissues and patients. For indications with small patient populations or for complex diseases demanding fine-grained stratification, statistical models must draw on these niche biological states to achieve translational utility.

Our first 30M cell atlas represents thousands of hours of human labor - the combined efforts of the Pythia Biosciences, Miraomics and LatchBio curation teams. It spans broad space of relevant biology, with 150 diseases, 200 tissues, 27 technologies.

This atlas is available in a public portal. Customers can search by study title, abstract or controlled ontology terms and immediately download their data with transparent pricing.

A Partnership with Data Focused Solution Providers

LatchBio is a platform for solution providers. Our team builds compute infrastructure, tools, delivery portals and white labeling features so our customers can focus on their strengths: developing new assays, kits and services.

Our partnership with Miraomics and Pythia Biosciences expands our focus to data delivery. They leverage their expertise in data curation. We provide them with new technology to increase curation throughput, quality and distribute their services to more customers.

An Agentic Human-in-the-Loop Framework for Curation

Recent advances in foundation models and agentic workflows show promise in autonomous scientific reasoning and software development. We hypothesized the structure of the curation problem was particularly suited for these emerging technologies.

We’re also introducing latch-curate, an agentic Python framework that guides an expert scientist through an ordered, step-by-step curation lifecycle and helps them perform tasks like count matrix construction, cell typing and metadata harmonization with greater efficiency and accuracy.

A detailed description of the system design and function can be found in this whitepaper. Reach out to our team for access to this framework.

Our team is building a lot of interesting product at the intersection of molecular data curation and delivery. If you are a solution provider interested in trying our new tools or hosting your data on our portal, please reach out to kenny@latch.bio

Breakthroughs like Curio Bioscience's Seeker and Trekker capture spatially-resolved whole transcriptomes by placing tissue sections on barcoded surfaces. Vizgen's MERFISH uses sequential rounds of single-molecule fluorescence in situ hybridization with combinatorial labeling for each RNA species. NanoString's CosMx Spatial Molecular Imager can detect 18,000 genes simultaneously at subcellular resolution.

The ability to construct complete spatial and temporal atlases of entire organs across disease states and developmental stages is finally attainable. But these new technologies bring a new scale and complexity of data. Individual experiments routinely produce terabyte datasets, combining high-resolution microscopy images, spatial coordinate mappings, and molecular count matrices.

Each platform outputs proprietary data structures and formats, with preprocessing and data wrangling before integration with open-source visualization tools. Existing viewers struggle with performant image and molecular scatterplot rendering, creating a frustrating user experience hardly conducive to free biological exploration. Beyond visualization, scientists must toggle between viewing data and conducting sophisticated analyses: differential gene expression, cell-type deconvolution, ligand-receptor analyses, and more.

Much like early genomics required genome browsers to contextualize high-throughput data, spatial biology now demands an analogous tool: one that supports dynamic, multi-modal and interactive visualization at the scale of millions of cells, all while plugging into a robust ecosystem of ready-to-use, open-source tools.

Here we introduce an H5 viewer for large spatial data built from new systems technologies and outline its design, function and practical use in real biological flows.

Design Principles

The H5 viewer is a widget in Latch Plots: a Python-based, reactive notebook purpose built for high-throughput biological data analyses.

from lplots.widgets.ann_data import w_ann_data

w_h5(ann_data=adata, spatial_dir="latch://117.account/some-data")

The widget takes two parameters: (1) an AnnData object and (2) a directory that contains multiple image layers - morphology, nuclear stains, H&E scans - and a DuckDB database of transcripts.

Each image layer is converted into a PMTile file: a pyramidal format of hierarchical image tiles originally built for rendering interactive maps. The format allows fast, random access to image chunks through HTTP range requests directly from blob storage. This allows smooth zooming, panning, and multi-layer visualization in the browser without a dedicated backend server.

Our team repurposed the technology to performantly stream chunks of high content spatial images to molecular viewers and wrote code to build these files directly from the outputs of popular spatial vendors, like CosMx, Singular and Xenium.

A DuckDB database file with transcript coordinates and metadata is stored in this same directory. DuckDB is a lightweight, in-memory database that allows direct access to large transcript data with a performant SQL engine.

The combination of AnnData, PMTiles and DuckDB allow the precise and low-latency overlay cell, transcript and image information in real time. Researchers can filter transcripts by gene name, region, expression level or cell type while rendering relevant sections of high resolution morphology.

A Bet on H5

While standards for spatial data storage are still fragmented, we are seeing convergence around the H5 file format. Originally developed for early high dimensional data from satellite imagery and particle colliders in the early 2000s, HDF5 (H5) first gained widespread adoption in biology when single cell techniques began producing abundant count matrices around 2010.

The H5AD format, a precise way of storing transcripts and metadata in an H5 file, was popularized by the AnnData project and continues to gain adoption with native Python libraries and abundant community driven best practices. As “higher order” molecular assays began supplementing transcript counts with new measurements - like images, proteins and epigenetic signatures - teams started storing this information in or around the core count matrix. With strong libraries, filesystem-like internals and community inertia, we’ve chosen to build our viewer around this format.

Adaptable to any Imaging and Sequencing Instrument

Because each spatial vendor generates different output files, our team wrote a Python library that constructs spatial AnnData objects, PMTiles files and DuckDB files directly from native instrument outputs. These files should “just work” with the H5 viewer.

The library supports widely used platforms, like Nanostring CosMx, 10X Genomics Xenium, Singular G4X, and Vizgen MERFISH. We are eager to work with new companies and expand the library to new technologies.

Features

Color by Categories or Genes of Interest

In the viewer, you can color by genes of interest or various categories in the observation layer of the AnnData object.

Lasso-select and annotate

Use the lasso selection tool to select clusters of cells and assign them a unique label. Any changes to the UI automatically mutate the underlying AnnData object so it can be used elsewhere in the analysis within the same notebook.

Image Alignment

Spatial technologies often capture brightfield or H&E images from the same slide or serial sections. The H5 viewer supports standard formats (TIFF, PNG, JPEG) and overlays them onto AnnData spatial embeddings using a landmark-based coordinate mapping system.

The viewer offers two alignment modes: affine alignment and STAlign.

Affine alignment applies a linear transformation matrix to account for translation, rotation, scaling, and shear. It requires at least three corresponding landmarks and uses least squares optimization to minimize residuals, making it well-suited for minimally deformed tissues.

In contrast, STAlign is designed for cases where tissue deformation occurs, such as with serial sections in Slide-seq, DBiT-seq, or Stereo-seq. It performs non-rigid registration by locally warping the image to match the spatial coordinates of the transcriptomic data, allowing for alignment even when the tissue is stretched, folded, or distorted.

While STAlign typically requires a multi-step setup (loading data into a Jupyter notebook, entering precise X and Y landmarks, tuning regularization parameters, and optionally using a GPU for acceleration), the H5 viewer abstracts all of this. Users can interactively define correspondences through the interface, and the alignment runs in the background, delivering robust registration without requiring any code.

Scale to millions of cells

The same enormous and multi-layered data that make spatial techniques so powerful also make visualization a challenging engineering problem. A CosMx machine can output 464K cells, 543M transcripts and multiple layers of high content images. In aggregate, the file data can approach 1TB for a single experiment on 100mm^2 of tissue. This volume will only increase as assay throughput continues to follow linear (and in some cases exponential) curves.

Moving large precise and large quantities of data from disk, over network and into a browser requires experience with storage devices, file systems and browser internals. A deep understanding of systems will be required to continue writing performant software as instrument throughput continues to climb. (We encourage engineers interested in these problems to reach out to the team)

Dispatch Compute Intensive Workflows

Visualizing the images and molecules is just the start of spatial analysis. The true scientific value comes from doing things with the data. Researchers need to toggle between visual inspection and spatial domain detection, neighborhood interaction mapping, or pseudo-bulk differential expression across tissue regions.

Interactive Analysis Suite

Latch Plots provides direct access to essential analytical tools from the H5 viewer, including quality control filtering, PCA/UMAP recomputation, Leiden clustering, and differential expression analysis. Unlike static visualization platforms, the Python for these interfaces can be accessed and extended in the browser. The "Edit" mode reveals the underlying Python code for every UI component and any data transformation.

This flexibility enables deep customization: Latch engineers have collaborated with technology providers to modify plot templates with platform-specific modules, creating tailored workflows for distinct spatial biology technologies.

Compute-Image Integration

Advanced spatial analyses require tight coupling between tissue visualization and resource intensive tools. Ligand-receptor interaction analysis must toggle between spatial regions of interest and long workflows. Similarly, cell type deconvolution can take over a day on complex datasets and require some interaction with a workflow orchestrator.

Researchers use lasso selection tools within the H5 viewer to define cellular populations or spatial regions, then trigger compute-intensive analyses with a single "Launch" button. These workflows are defined via Latch's Python, Nextflow, or Snakemake SDKs, providing users with real-time progress monitoring and complete execution logs. Upon completion, results are automatically integrated back into the Plots environment for immediate visualization and downstream analysis.

Scalable Ecosystem

Currently, Latch supports over 12 downstream analysis modules, opening the gateway to a robust infrastructure and ecosystem of ready-to-use, open-source tools.

Real Flows

The best way to understand a tool is to understand how it tackles real tasks. We’ll show the H5 viewer in action with a handful of realistic biological flows.

Takara Bio Seeker and Trekker: Spatial Temporal Atlas of Mouse Ovaries

Ovulation is essential for reproductive success, yet the underlying cellular and molecular mechanisms are far from clear. We reproduced a study from Mantri et. al (2024), which used Takara Bio’s Seeker technology to build a detailed, time-resolved atlas of mouse ovarian follicle maturation and ovulation.

Data Acquisition and Processing

First, raw FastQ files can be downloaded from GEO, along with Hematoxylin and Eosin (H&E)-stained images of tissue sections from mouse ovaries at seven distinct stages, spanning from preantral to post-hCG-triggered preovulatory phases.

Using Seeker’s hosted bioinformatics pipeline on the Latch Platform, FastQ files can be processed into multiple H5AD objects, one per sample. The Seeker pipeline on Latch also pre-computed the UMAP, ran SCTransform for normalization, and found spatially variable genes. All preliminary results are organized into a user-friendly HTML file that can be opened directly on Latch Data.

Sample Integration and Visualization

A dedicated workflow merged all per-sample H5AD files into a unified object, introducing spatial offsets, which enabled side-by-side sample visualization in the H5 viewer. This integration facilitated comparative analysis across developmental stages.

Quality Control and Analysis Pipeline in Plots

Users can then open a Latch Plots notebook, where they can ingest the H5AD object and open it in the H5 viewer. In the same notebook, we performed quality control (QC), filtering, identified highly variable and spatially variable genes, and applied Leiden clustering.

Spatial Segmentation and Image Alignment

Spatial data analysis requires precise segmentation of tissue regions based on morphological features. We examined spatial embeddings for individual samples, uploading corresponding H&E images to the H5 viewer for detailed examination. The viewer's interface provided intuitive controls for image manipulation, including rotation, zoom, pan, and flip functions.

A unique aspect of Seeker technology involves capturing H&E images from serial tissue sections rather than identical slides. This approach introduces alignment challenges, as tissue sections may exhibit distortions, shearing effects, and imperfect overlap that cannot be resolved through simple affine transformation methods.

To overcome these limitations, H5 viewer comes with an interactive anchor point assignment system and STAlign, a GPU-accelerated alignment tool. STAlign addresses the complexities of partially matched tissue sections and local non-linear distortions through diffeomorphic metric mapping, providing robust alignment solutions for challenging tissue geometries.

Follicle Identification and Downstream Analysis

Using aligned images as reference, we lasso selected and labeled follicles within each sample. These follicle regions serve as pseudo-replicates for subsequent pseudo-bulk analyses using the DESeq2 pipeline on Latch, enabling the identification of distinctive gene expression signatures across different cell types throughout follicular development.

Cell Type Deconvolution and Ligand-Receptor Analyses

Each 10-micron Seeker bead captures transcripts from approximately 1–2 cells, making cell type deconvolution necessary. In Latch Plots, the transcriptomic H5AD can be saved to Latch Data and processed using the cell2location workflow.

The resulting annotated H5AD file can then be reloaded into Plots and visualized in the H5 viewer to inspect predicted cell types.

Once we were confident in cell-type assignments, we performed ligand-receptor analysis. Squidpy’s ligand-receptor module is usually computationally intensive due to the large number of spatial permutations it evaluates. To optimize performance and relevance, we utilized the lasso-selection tool in the H5 viewer to define specific regions of interest. This focuses the analysis on cell-cell interactions that are spatially proximal, filtering out biologically implausible long-range interactions.

AtlasXOmics: Spatial Epigenomics Analysis

Epigenetic profiling assays are essential for understanding gene regulation, but traditional approaches lack spatial context. Spatial epigenomics reveals how chromatin modifications distribute across native tissue architecture, illuminating the geometric relationships between regulatory states and disease biology.

AtlasXOmics developed deterministic barcoding in tissue sequencing (DBiT-seq), the first platform enabling spatially-resolved epigenome profiling. Using microfluidic chips to deliver barcodes to precise tissue coordinates, DBiT-seq extends single-cell ATAC-seq and CUT&Tag assays to spatial contexts, uncovering causative relationships between chromatin accessibility, histone modifications, and context-specific gene expression within intact tissues.

However, synthesizing tissue morphology with epigenetic state relies on techniques in image processing and sequencing bioinformatics that make data analysis challenging.

Data Processing

ATAC-seq generates a large files of barcoded sequencing reads, each of which corresponds to a unique tranposase insertion event at the sequenced locus. These reads must be aligned to a reference genome, their barcodes error corrected, their adapters trimmed and aligned coordinates adjusted for known behavior of the transposase (Tn5-shift). The result is a fragment file, annotating sequenced fragments with genomic coordinates and error corrected barcodes.

To do this, AtlasXOmics engineers developed a chromap pipeline using the Latch Python SDK, which defines the system requirements, dependencies, and compute resources, as well as a graphical interface enabling scientists to process this data directly without bottlenecking their cadence of experimental work. The pipeline can be run in batch mode to generate fragment files for multiple samples in parallel.

Peak Calling and Differential Peak Analysis

A second AtlasXomics workflow processes fragment files and sample metadata to perform MACS2-based peak calling, peak-to-gene annotation, motif enrichment analysis, and differential accessibility quantification across experimental groups.

The workflow generates three specialized H5AD files for downstream analysis. The cell-by-gene matrix aggregates chromatin accessibility signals at the gene level. The cell-by-motif matrix captures transcription factor activity deviations across spatial regions. The pseudobulked matrix summarizes group-level accessibility patterns, supporting differential analysis for region-specific regulatory signatures.

Visualizing H5AD files with the H5 viewers

These files can be opened directly in a Latch Plots notebook and visualized using the H5 viewer. If an aligned H&E image is available, it can be overlaid onto the spatial coordinates in the H5AD file to contextualize chromatin accessibility or motif deviations within tissue morphology.

In the same Plot notebook, AtlasXOmics engineers developed analysis modules for comprehensive spatial epigenomics interrogation.

Gene Accessibility Analysis

The gene accessibility viewer shows spatial scatter plots of gene accessibility across tissue. The heatmap clusters cells and genes by accessibility profiles. Volcano plots identify differentially accessible genes across groups, while the ranked score plots order genes by effect size or significance.

Motif Activity Mapping

The motif scatter displays spatial maps of motif accessibility deviations. The heatmap summarizes deviation scores across clusters. Volcano plots highlight condition-specific motifs, and ranked score plots prioritize motifs by regulatory impact.

Neighborhood Enrichment

Another tab performs spatial enrichment of gene or motif signals within selected regions, enabling comparison of localized regulatory states.

Genome Browser Integration

AtlasXOmics team added an genome-browser widget in Latch Plots to display fragment alignments and peaks. This can be used to compare peaks and chromatin accessibility across samples or conditions.

Exciting future directions include the ability to enable lasso selection of tissue regions on H&E images, automated generation of bigWig files via Plots cells, and direct ingestion into the IGV viewer for comparative peak analysis between spatial regions.

Together, the H5 viewer and its supporting modules enable fast, interactive interrogation of spatial epigenomic data, linking molecular state, tissue morphology, and regulatory activity in one interface.

Singular: Multi-omic Analysis (Proteome, Transcriptome, H&E) with G4X

The Singular’s G4X™ Spatial Sequencer includes a gene panel derived from single-cell transcriptomic sequences and an accompanying protein panel. In addition, it offers H&E imaging. G4X pushes the boundary by bringing spatial, FFPE-compatible, multi-omic sequencing onto the same benchtop box that already performs bulk NGS, closing several of the biggest gaps left by traditional CITE-seq while trading off maximum protein depth for spatial context.

Viewing Multiple Modalities in One Viewer

With this technology uniting multiple biological modalities, we developed a downstream analysis dashboard marrying information from all data sources in one place.

The H5 image viewer enables scientists and pathologists to examine transcriptomic and imaging data simultaneously, facilitating the easy overlay of transcriptomic and proteomic data on H&E images.

Joint RNA and Protein Clustering

After normalizing and transforming the transcript counts and protein intensity values, we learned a joint clustering from the transcript-count and protein-intensity data. We observed that cluster quality improved when both data modalities were incorporated.

Identification of differentially expressed genes

We then identified cluster 2 and found that the IGHG1 transcript was differentially expressed in this cluster versus the others. Moreover, we saw a co-expression of cytoplasmic antibodies in the cells differentially expressing IGHG1, a canonical signal for plasmablasts and mature plasma cells.

Because the H5 viewer is implemented as a widget, users can instantiate it multiple times within a notebook, enabling side-by-side comparisons of different tissues, conditions, or time points.

To Conclude

The increasing size and complexity of spatial data has created one of the most challenging human computer interaction problems in the history of computing. Our H5 viewer tackles challenging problems in data intensive engineering, consolidates fragmented spatial machine outputs into standard inputs and brings together a rich ecosystem of open source tertiary analysis tools. Our team is excited to continue to improve these tools as spatial continues to progress and expand support for new technologies.

LatchBio works directly with many of the spatial vendors described here, like Curio and AtlasXOmics, providing white labeled infrastructure and analysis tools that help their customers close the loop on analysis and sell more kits. If you are a spatial solution provider interested in providing these tools to your customers, reach out.

It’s easy to get swept up in the scale and ambition of today’s single-cell efforts. But none of it would be possible without the steady innovation from kit providers and technology developers. They’ve pushed the limits on how many cells we can measure and how deeply we can profile each one. Kit providers may not sound glamorous, but they are modern-day van Leeuwenhoeks and Hookes. Like the inventors of the microscope, they gave us our first clear view into the cell’s inner workings.

Over the past decade, 17 companies and 61 (!) kits have emerged to interrogate the transcriptional states of a single cell.

To help readers navigate the landscape, we are releasing a live database of all existing single-cell kits and technologies at https://latch.bio/spatial-single-cell-landscape?type=single-cell

This is an extension of our previously published spatial technologies market map, which featured 20 companies and 42 products defining spatial biology today.

What’s Covered

In this article, we walk through the landscape of existing commercial single-cell technologies.

We will cover:

• Key Technologies
  How do you isolate or identify a single cell?

• Key Chemistries
  How do you read the mRNA from the single cell?

• Methods by Attributes: Comparisons following criteria:

  • Number of Cells Supported
    High-, mid-, and low-throughput technologies

  • Analyte Support
    Whether the platform supports RNAs only or includes DNA, proteins, or epigenetic markers.

  • Multiomics Support
    Support for capturing multiple analytes from the same cell.

  • FFPE Compatibility
    Whether the technology can be applied to formalin-fixed paraffin-embedded samples.

  • Cloud Bioinformatics
    Whether vendor-provided software is included to analyze and visualize results.

Part 1: How do you isolate a single cell? Key approaches

Single-cell isolation is the first step in most single-cell sequencing workflows. There are four primary technology groups used to achieve this, each with distinct mechanisms and trade-offs:

1. Capsule-based Methods
   These approaches use microfluidic droplets or capsules as miniature reaction chambers to isolate individual cells. Each capsule contains the necessary reagents for downstream processing, such as barcoding and lysis.

   • Key Platforms:

     • 10X Genomics Chromium

     • Atrandi Biosciences Semi-Permeable Capsules

     • Bio-Rad ddSEQ

2. Microwell-based Methods
   Microwell systems rely on physical wells small enough to capture a single cell. Some platforms include integrated imaging systems to confirm the presence of a single cell in each well before proceeding to library prep.

   • Key Platforms:

     • Takara Bio Shasta

     • Singleron GEXSCOPE® chip

     • BD Rhapsody

3. Combinatorial Indexing Methods
   Instead of physically isolating single cells, these methods rely on repeated rounds of molecular barcoding across split-pool steps. Cells or nuclei are indexed through successive rounds of barcode addition, enabling single-cell resolution without compartmentalization.

   • Key Platforms:

     • Parse Biosciences Evercode

     • Scale Biosciences Quantum Barcoding technology

4. Kinetic Confinement Method (pioneered by CS Genetics): Paramagnetic beads carrying barcodes bind to individual cells, and upon heat activation in a viscous buffer, cell lysis and barcode capture occur in a confined microenvironment, enabling high-fidelity single-cell indexing.

5. Matrigel-based Method: A less common approach involves embedding cells in a hydrogel-like matrix (e.g., Matrigel), then performing reactions in situ. This method saw limited adoption and is no longer active commercially.

   • Key Platform:

     • Scipio Bioscience Asteria™ (Company ceased operations in September 2024; included here for completeness)

We walk through each method and the commercial kits that implement them in the sections below.

Capsule-based Methods

In capsule-based approaches, each single cell is enclosed in a physically isolated compartment that enables independent biochemical processing. Within this capsule, cell lysis releases RNA, which undergoes reverse transcription and cDNA synthesis using reagents that include unique molecular identifiers (UMIs) and cell barcodes. The compartment ensures that the barcoded cDNA generated corresponds to a single cell.

10x Genomics

Instrument required: Yes

10x Genomics uses microfluidic droplet generation to co-encapsulate single cells and barcoded gel beads within oil emulsions. Each gel bead is functionalized with oligonucleotides containing a cell-specific barcode and a poly-dT sequence that captures mRNA. Once inside the droplet, the cell is lysed, and mRNA is hybridized to the bead’s oligos. Reverse transcription occurs within the droplet, producing barcoded cDNA tagged with both UMI and cell barcode.

Atrandi Biosciences

Instrument required: Yes

Atrandi introduces semi-permeable capsules (SPCs) formed from hydrogel polymers. Each SPC encapsulates a single cell and is engineered to allow diffusion of small molecules such as enzymes and nucleotides, while retaining larger macromolecules like RNA and DNA. Cells undergo lysis inside the capsule, followed by in-capsule reverse transcription and amplification. The porous structure allows reactions to be performed in bulk solution without cross-contamination, making the workflow highly scalable.

Bio-Rad ddSEQ

Instrument required: Yes

The Bio-Rad ddSEQ system combines microfluidics with droplet partitioning. Cells are individually encapsulated into nanoliter droplets containing barcoded primers immobilized on beads. Each droplet functions as a miniature reaction chamber, where cell lysis, mRNA capture, and reverse transcription take place. The barcoded beads encode both a unique cell barcode and UMI, allowing for downstream demultiplexing and transcript quantification. The ddSEQ platform is particularly suited for targeted gene panels and integrates with Illumina sequencing workflows.

Mission Bio Tapestri

Instrument required: Yes

Mission Bio’s Tapestri platform performs targeted single-cell DNA and protein analysis using a two-step microfluidic droplet workflow. In the first step, individual cells are encapsulated into droplets along with lysis buffer and protease, releasing genomic DNA and surface proteins. These lysates are then recovered and re-encapsulated in a second droplet step with barcoded primers targeting specific genomic loci. Within these droplets, PCR-based amplification captures genetic variants alongside cell-identifying barcodes. Surface protein expression is simultaneously measured via antibody-oligo conjugates included in the workflow.

This two-step separation of cell lysis and barcoding allows for optimized conditions at each stage and ensures high specificity in variant detection. The Tapestri platform enables simultaneous profiling of single-nucleotide variants (SNVs), insertions and deletions (INDELs), copy number variations (CNVs), and protein expression from the same single cell, providing multi-omic resolution ideal for applications in cancer genomics and cell therapy development.

Illumina (previously Fluent PIP-seq Kit)

Instrument required: No

Illumina’s PIPseq™ Single-Cell RNA Prep, originally developed by Fluent BioSciences, employs a microfluidics-free, vortex-based method to partition individual cells into hydrogel-based capsules known as Particle-templated Instant Partitions (PIPs). By simply vortexing a mixture of cells, barcoded beads, and reagents, the system forms emulsions that encapsulate single cells without the need for specialized instruments. Within each PIP, cell lysis, mRNA capture, and reverse transcription occur, utilizing barcoded oligonucleotides to tag transcripts uniquely per cell

Microwell-based Methods

Singleron: SCOPE® Technology

Instrument required: No

The SCOPE® platform uses a custom-designed PDMS chip containing thousands of uniformly sized microwells etched into the surface. Cells are introduced by gravity and settle into the wells, which are preloaded with barcoded oligonucleotides. The chip is then sealed with a reagent-containing cover to initiate cell lysis and reverse transcription. The chip layout ensures minimal cross-talk between wells and allows for efficient reagent diffusion. Its geometry supports high capture efficiency and is optimized for “one-pot” reactions, reducing sample handling and transfer loss.

TakaraBio

Instrument required: Yes

The Shasta Single Cell System from Takara Bio is a high-throughput platform designed for comprehensive single-cell genomic and transcriptomic profiling. It integrates automated nanoliter-scale dispensing, optical imaging, and cell selection to isolate and process individual cells for next-generation sequencing.

The system supports total RNA sequencing with full gene-body coverage (Shasta Total RNA-Seq), enabling the detection of complex transcriptomic features such as splicing isoforms, gene fusions, and segmental copy number variations. For single-cell DNA applications, Shasta offers a whole-genome amplification module (Shasta WGA) to profile genome-wide variation at the single-cell level.

An integrated fluorescence imaging system allows for identification and characterization of stained cells, while the CELLSTUDIO software enables assay customization and optimization.

Paired with Cogent NGS tools for downstream analysis, the Shasta system offers a modular, end-to-end workflow suited for discovering transcriptional and genomic heterogeneity across thousands of single cells.

BD Biosciences: Rhapsody™

Instrument required: Yes

The BD Rhapsody™ system is based on a microwell cartridge that contains tens of thousands of tightly packed wells arranged in a high-density grid. Each well is engineered to capture a single cell and a single barcoded magnetic bead, enabling localized reactions for mRNA capture. The beads are functionalized with oligonucleotides containing cell barcodes, unique molecular identifiers (UMIs), and poly-dT sequences for hybridizing to polyadenylated RNA.

Once cells are loaded by gravity and settle into the microwells, lysis buffer is introduced directly onto the cartridge. The released mRNA hybridizes to the barcoded oligos on the beads. After hybridization, beads are magnetically recovered and pooled for reverse transcription and library preparation.

BD offers two versions of the Rhapsody system to accommodate different throughput needs:

• BD Rhapsody Express:
  A benchtop instrument that processes a single microwell cartridge per run. This format is ideal for moderate-throughput labs and supports flexible experimental design with integrated fluidics and imaging for quality control.

• BD Rhapsody HT Xpress:
  A high-throughput variant that can process up to 8 cartridges in parallel. This system is designed for large-scale studies, offering substantial gains in cell throughput, reagent batching, and hands-off operation. The HT configuration enables processing of up to 160,000 cells per run (assuming 20,000 cells per cartridge), making it suitable for cohort-scale immune profiling or large multiplexed experiments.

Flexomics: FLEX™ Chip

Instrument required: No

The FLEX™ chip is a microscope-compatible picowell array designed to capture and isolate individual cells in a spatially resolved format. Each picowell physically traps a single cell, allowing researchers to perform cell-based assays directly on-chip. The chip also supports live-cell imaging, functional perturbations, and phenotypic assays without the need for sample transfer.

After cellular assays are completed, the same chip is used for molecular analysis such as transcriptomics or proteomics. Molecular barcoding reagents are added to capture both mRNA and oligo-tagged antibodies within each well. Cell-of-origin information is preserved from phenotypic imaging through to sequencing because each cell remains in a fixed spatial location throughout the workflow.

Celldom: CloneXplorer Platform

Instrument required: Yes

CloneXplorer is Celldom’s next-generation live single-cell assay platform that combines continuous functional imaging, molecular profiling, and automated cell retrieval in a single, high-throughput system. Built on a microfluidics-free design, CloneXplorer uses a standard 96-well plate format, where each well contains thousands to tens of thousands of microwells, allowing simultaneous analysis of over 100,000 single cells in one experiment.

The platform enables extended, real-time observation of single-cell dynamics through timelapse 4-color imaging, capturing morphological, functional, and molecular behaviors at high resolution. This capability is particularly powerful for applications such as CAR T-cell potency testing, monoclonal antibody (mAb) or TCR discovery, clonal selection for cell line development, and precision oncology. AI-powered cell detection and tracking algorithms analyze each cell’s behavior and link functional outputs to downstream sequencing-based molecular profiles. CloneXplorer also includes an on-instrument picker, which allows researchers to isolate specific cells or clones of interest for further expansion or analysis.

BioSkryb

Different from BD and Singleron, BioSkryb platforms use standard 96- and 384-well plates, modified for single-cell compatibility. The well design allows for direct sorting of single cells via FACS or automated liquid dispensers. Each well serves as a discrete reaction chamber for PTA-based whole genome amplification or RNA processing.

Combinatorial-Indexing Methods

Combinatorial indexing enables single-cell resolution without physically isolating each cell. Instead, cells or nuclei are processed in bulk, and unique barcodes are introduced through multiple rounds of molecular tagging. Each round involves pooling, splitting, and re-barcoding the sample, generating statistically unique barcode combinations corresponding to individual cells. This approach eliminates the need for microfluidics or specialized chips and scales efficiently to hundreds of thousands to millions of cells per run.

Scale Biosciences: Quantum Scale Technology

Scale Biosciences' Quantum Barcoding™ platform employs a three-round barcoding strategy optimized for speed and simplicity:

1. Reverse transcription: In the first round, cells or nuclei undergo reverse transcription, adding an RT barcode and a molecular barcode to each transcript (Result: cDNA of Transcript X + RT barcode + Molecular barcode). All cells are pooled and redistributed to the second plate, the Quantum Barcoding Plate.

2. Quantum Barcoding: In the second round, barcoded cDNA is ligated to beads within a Quantum Barcoding Plate, each well of which contains a unique barcode (Result: Microwell barcode + previous cDNA of Transcript X + RT barcode + Molecular barcode). All wells are then pooled and redistributed into tubes.

3. PCR indexing and final library construction: Library indices are added during PCR amplification.

This plate-based ligation replaces the need for multiple split-pool steps and significantly reduces hands-on time. The resulting barcode combinations uniquely tag transcripts from each single cell. This method supports up to 675 million unique barcode combinations, making it well suited for profiling millions of cells in a single run. The streamlined, low-touch protocol is also compatible with automated liquid handling.

Parse Biosciences: Evercode Technology

Parse’s Evercode™ platform uses a four-round combinatorial barcoding strategy to uniquely label RNA transcripts from individual cells or nuclei.

1. Reverse transcription:

   1. Split: Samples are distributed into wells and the first, sample-specific barcodes are applied to fixed cells or nuclei with an in-cell reverse transcription (RT) reaction.

   2. Pool: Cells from each well are pooled together.

   3. Result of first step: Cell + Sample-specific barcode

2. Ligation:

   1. Split: Cells or nuclei are distributed across a plate and an in-cell ligation appends the second barcode.

   2. Pool: Cells from each well are pooled together.

   3. Result of 2nd step: Cell + Sample-specific barcode from 1st step + Second barcode from in-cell ligation

3. Ligation

   1. Split: The third barcode is applied with another in-cell ligation after the cells or nuclei are split across a plate.

   2. Pool: Cells from each well are pooled together.

   3. Result of 3nd step: Cell + Sample-specific barcode from 1st step + 2nd barcode from in-cell ligation + 3rd barcode from in-cell ligation

4. Lysis + PCR:

   1. The pooled cells are divided across several sublibraries. The cells are lysed and the fourth, sublibrary-specific barcode is applied by PCR.

   2. Result of 4th step: Cell + Sample-specific barcode from 1st step + 2nd barcode from in-cell ligation + 3rd barcode from in-cell ligation + 4th barcode applied by PCR.

5. Library Prep: The library preparation appends adapters ready for loading on any next-generation sequencer.

One of the unique features of Parse Biosciences kit is cells can be fixed and stored up to 6 months, supporting flexible batching and processing across days. The system enables profiling of up to 5 million cells and 384 samples per experiment without requiring specialized instrumentation.

Kinetic Confinement Method

CS Genetics: SimpleCell™ platform

CS Genetics creates its own category, offering a unique method known as called Kinetic Confinement (KC).

KC operates using two key components:

• CPair Beads: These are paramagnetic beads roughly the size of mammalian cells. Each bead is coated with unique indexing oligonucleotides (comprising a poly-dT sequence and a cell-specific barcode) and antibodies that enable general cell binding.
  
  

• Kinetic Confinement Buffer (KCB): This is a thick, viscous solution that limits molecular diffusion and includes a heat-activated lysis agent. It also acts as a cryoprotectant, allowing samples to be stored and processed later.

In practice, cells are combined with the CPair beads, then KCB is added. Heating the mixture triggers both cell lysis and the release of oligonucleotides from the beads. This creates a concentrated microenvironment around each bead where mRNA molecules and barcodes are in close proximity, enabling efficient hybridization.

For reverse transcription, second-strand synthesis is initiated using randomly distributed primers instead of traditional 5’-anchoring methods. This design eliminates the need for unique molecular identifiers (UMIs). Instead, each transcript is marked by the random position at which second-strand synthesis begins, termed the “5S” site. Since it's statistically improbable for two independent transcripts to start at the same site, duplicates can be confidently identified as PCR artifacts, leading to cleaner, more accurate data.

Part 2: How do you read mRNA? Key Chemistries:

Refresher on the Direction of RNA

In both RNA and DNA, each strand has a built-in direction, determined by the structure of its sugar-phosphate backbone. This directionality is defined by the numbering of carbon atoms in the ribose (or deoxyribose) sugar:

• The 5′ end refers to the fifth carbon, where a phosphate group is typically attached.

• The 3′ end refers to the third carbon, which carries a free hydroxyl (-OH) group.

This 5′ to 3′ orientation is critical for how genetic material is copied, transcribed, and read.

DNA can only be synthesized in vivo in the 5′-to-3′ direction. Here’s why:

Incoming nucleotides arrive as triphosphates (e.g., ATP, GTP), which contain high-energy bonds between their phosphate groups. The hydroxyl group (-OH) at the 3′ end of the existing strand acts as a nucleophile, an electron-loving group that seeks out positively charged or electron-deficient atoms. It performs a nucleophilic attack on the α-phosphate of the incoming nucleotide, which is electron-rich due to the surrounding negatively charged oxygen atoms, but structurally set up to be reactive due to the strain in the triphosphate bond.

This reaction breaks off two phosphate groups (as pyrophosphate) and forms a phosphodiester bond, extending the DNA strand by one nucleotide.

This chemical mechanism is unidirectional. There’s no equivalent reactive group on the 5′ end, so all DNA and RNA synthesis proceeds exclusively in the 5′ → 3′ direction.

Why does directionality matter for RNA sequencing?

The 5′ to 3′ structure of RNA directly shapes how we design sequencing methods. Because all polymerases extend RNA or DNA by adding nucleotides to the 3′ end, reverse transcription (converting RNA into cDNA) naturally begins at the 3′ end of the RNA molecule.

Modern single-cell RNA-seq protocols take advantage of this property by anchoring capture primers at the polyadenylated (polyA) tail, a common feature of mature mRNA, allowing selective and efficient initiation of cDNA synthesis. This foundational principle is what underlies 3′ chemistry, one of the most widely used approaches in single-cell transcriptomics.

3’ Chemistry

3′ chemistry in single-cell RNA sequencing refers to protocols that capture and sequence only the 3′ end of polyadenylated RNA transcripts.

3′ chemistry relies on oligo-dT primers to hybridize to the poly-A tail of mRNA molecules. These primers are often part of a larger molecule (such as the GEM-X 10X 3’ Kit’s Gel Bead above) that also includes:

• A cell-specific barcode (to identify which cell the RNA came from)

• A UMI (to deduplicate PCR artifacts and count original transcript molecules)

• A PCR handle or sequencing adapter

The structure allows all mRNA transcripts from a single cell to be reverse-transcribed and tagged with the same barcode and different UMIs.

In 3′ single-cell RNA-seq chemistry, individual cells are isolated (e.g., in droplets or wells) and lysed to release mRNA, which is captured via oligo-dT primers that hybridize to the poly-A tails. These primers are pre-attached to a scaffold containing a cell barcode and a unique molecular identifier (UMI), enabling transcript origin and quantification. Reverse transcription is then performed, generating cDNA fragments anchored at the 3′ end of transcripts. All cDNA is pooled and PCR-amplified using universal adapters, and libraries are sequenced such that Read 1 captures the barcode and UMI, while Read 2 sequences a short region adjacent to the 3′ end of the transcript.

This 3′ capture approach has several benefits. First, by targeting the poly-A tail present on nearly all mature mRNAs, it enables broad transcriptome coverage without needing gene-specific primers or probes. This means any polyadenylated RNA can, in principle, be captured, offering wider, unbiased capture of the transcriptome. Second, because only the region near the 3′ end is reverse-transcribed and sequenced, the resulting libraries are small, and short reads (~100 bp) are sufficient to map the transcript to a gene. This significantly reduces sequencing costs per cell, enabling higher throughput, often profiling tens of thousands of cells to hundreds of thousands in a single experiment.

Examples of 3’ Kits:

• 10x Genomics: Chromium Single Cell 3′ v3.1

• Parse Biosciences: Evercode Whole Transcriptome

• BD Biosciences: Rhapsody WTA

• Singleron: GEXSCOPE® scRNA-seq

• Flexomics: FLEX™ RNA + Protein

However, this method cannot reliably resolve transcript isoforms, splicing events, or full-length gene structures. These features typically involve differences in internal or 5′ exons, alternative splice sites, or variable transcription start sites, all of which are located far upstream of the 3′ end.

Full-length or 5′ capture strategies are better suited for applications requiring this level of resolution.

5’ Chemistry

5′ chemistry in single-cell RNA sequencing refers to protocols that capture and sequence the 5′ end of polyadenylated RNA transcripts. This strategy is particularly valuable for studying transcription start sites and for recovering full-length immune receptor transcripts (e.g., TCR and BCR) necessary for V(D)J repertoire profiling.

Reverse transcription in 5′ chemistry still initiates at the 3′ poly-A tail, using an oligo-dT primer that also includes a cell barcode, UMI, and a partial sequencing adapter. As reverse transcriptase synthesizes cDNA toward the 5′ end of the mRNA, it eventually reaches the 5′ cap structure, a modified guanine nucleotide linked via a 5′–5′ triphosphate bridge. This cap is recognized by the reverse transcriptase’s terminal transferase activity, which adds several untemplated cytosine residues (typically CCC) to the 3′ end of the newly synthesized cDNA.

A template-switching oligonucleotide (TSO), designed with a 3′ GGG motif, hybridizes to these added cytosines and acts as a new template. The reverse transcriptase then extends the cDNA using the TSO sequence, effectively appending a universal adapter to the 5′ end of the cDNA. This completes a fully functional cDNA molecule flanked by known adapter sequences at both ends.

This structure is critical for downstream PCR and sequencing: it ensures that Read 2 starts at the 5′ end of the original mRNA, enabling accurate capture of transcription start sites and full-length variable regions of TCR/BCR transcripts.

Examples of 5’ Kit:

• 10x Genomics: Chromium Single Cell 5′ v2 + Immune Profiling V(D)J Enrichment Kits

• Singleron: GEXSCOPE® V(D)J and TCR/BCR Kits

In addition to 5’ cDNA synthesis, both kits use targeted enrichment of BCR/TCR regions. This can be done using pools of forward primers that bind to conserved regions at the 5′ end of V gene segments, paired with reverse primers targeting the constant (C) region. This selective amplification enables high-sensitivity recovery of full-length immune receptor transcripts, including the highly diverse CDR3 region, which is essential for reconstructing clonal lineages and profiling immune repertoires at single-cell resolution.

Probe-Based Targeted Gene Panels

Probe-based single-cell RNA-seq chemistries enable highly sensitive and cost-efficient profiling of predefined gene sets. Instead of sequencing the entire transcriptome, these methods selectively enrich for transcripts of interest using hybridization or ligation-based probes. This targeted enrichment allows deeper coverage of low-abundance transcripts while dramatically reducing the required sequencing depth per cell.

In probe-based workflows, enrichment occurs either before or after barcoded cDNA synthesis, depending on the platform. The core mechanisms include:

1. Hybridization-Based Enrichment:

• Biotinylated or capture-tagged DNA probes are designed to hybridize to complementary sequences on cDNA molecules. Once bound, target-probe complexes are pulled down using beads (e.g. streptavidin-coated magnetic beads), selectively enriching target transcripts.

• Used by: Parse Biosciences (Gene Select)

2. Ligation-Based Probes:

• Pairs of oligonucleotide probes hybridize to adjacent sequences on the target transcript. If both probes bind correctly, a ligase joins them to form a single, amplifiable DNA fragment. These ligated products are then barcoded and amplified.

• Used by: 10x Genomics (Gene Expression Flex Kit)

These strategies are often compatible with cell barcoding and UMIs, preserving single-cell resolution. Some platforms allow simultaneous enrichment of both targeted genes and the whole transcriptome or combine transcript and protein detection.

Example of Probe-Based Kit:

• 10x Genomics: Chromium Fixed RNA Profiling (Flex): Ligation-based probe pairs bind to fixed and permeabilized cells or nuclei. Upon successful hybridization, the pairs are ligated and extended during droplet-based barcoding in GEMs. Oligo-conjugated antibodies can be added before fixation to co-detect surface protein markers. It’s also the only product on the market that enables high-sensitivity targeted profiling from fixed, archived, or low-quality samples (like FFPE).

• Singleron: FocuSCOPE® Targeted Capture Kit: This kit uses customizable hybridization-based probes to enrich for specific genomic features in parallel with whole transcriptome profiling. It allows up to 10 user-defined regions, including gene fusions, SNVs, and viral RNA, making it a good candidate for oncology, virology, and mutation detection in clinical samples.

• Parse Biosciences: Gene Select: Uses custom or pre-designed hybridization-based probes are applied after Evercode Whole Transcriptome library prep to selectively enrich genes of interest. Library is hybridized to probes, and bound molecules are captured on magnetic beads. mmune1000 (~1000 genes for immune profiling) or user-defined custom panels is available. .

Full-length RNA Capture Chemistry

SMART-seq

Full-length RNA capture chemistries in single-cell RNA sequencing are designed to preserve and sequence the entire mRNA molecule, from the 5′ cap to the 3′ poly-A tail. Unlike 3′ or 5′ end-tagging methods, which capture only short regions of transcripts, full-length protocols reconstruct complete transcript structures. This enables analyses such as isoform discovery, alternative splicing, allele-specific expression, and transcript-level mutation detection.

The most widely used method for full-length capture is SMART (Switching Mechanism at the 5′ end of RNA Template) chemistry, developed by Clontech and commercialized by Takara Bio. The workflow involves:

1. Poly-dT priming to initiate reverse transcription at the mRNA’s 3′ poly-A tail.

2. As reverse transcriptase reaches the 5′ cap, it adds a short string of untemplated cytosines.

3. A template-switching oligo (TSO) with a 3′ GGG overhang anneals to these Cs, allowing the RT to continue and append a known sequence to the 5′ end of the cDNA.

4. The resulting cDNA, flanked by known sequences, is amplified in full using PCR. No early fragmentation occurs, and no barcode is needed because each cell is processed separately in its own well.

5. Amplified full-length cDNA is typically fragmented (e.g., by tagmentation) and sequenced across the full transcript body, enabling read coverage throughout exons, splice junctions, and UTRs.

Note: In SMART-Seq, reverse transcription and cDNA amplification are performed on physically separated cells (in wells or nanowells), eliminating the need for early barcoding. Each cell is processed independently, and indexing is introduced later during tagmentation or PCR. This contrasts with 10x Genomics' 5′ chemistry, where the cell barcode and UMI are pre-attached to a gel bead and introduced during reverse transcription. As a result, sequencing is anchored to either the 5′ or 3′ end of the transcript, depending on the kit, and captures only a short fragment adjacent to the barcoded priming site. This design supports high-throughput gene quantification and V(D)J profiling but does not preserve full transcript structure.

Example SMART-seq Kit:

• Takara Bio: SMART-Seq® v4

Singleron MobiusScope 3’ and 5’ Full-Length mRNA Reading via Circulization

In addition to SMART-based methods, MobiusSCOPE® by Singleron Biotechnologies introduces a new approach to full-length RNA capture through a 3′ and 5′ circularization strategy.

Traditional single-cell RNA-seq methods, such as 3′ or 5′ end-tagging chemistries, attach the cell barcode and unique molecular identifier (UMI) at only one end of the transcript, typically the 3′ poly-A tail. While sufficient for gene expression quantification, this design restricts sequencing to fragments adjacent to the barcoded end and prevents comprehensive recovery of internal transcript features. Once cDNA is randomly fragmented for sequencing, reads originating far from the barcode site cannot be linked back to their original molecule, making full-length transcript reconstruction infeasible.

MobiusSCOPE, developed by Singleron Biotechnologies, overcomes this limitation by using a circularization-based strategy to retain and trace both ends of the transcript. In this workflow, mRNA from each single cell is first captured using poly-T oligonucleotides attached to cell barcoding beads. Reverse transcription is then performed to generate cDNA, during which a unique barcode and UMI are incorporated at the 3′ end of each transcript. This ensures that every cDNA molecule carries a unique molecular tag that identifies its cell of origin and transcript identity.

To capture full-length information, the cDNA is then enzymatically circularized, bringing the 5′ end of the mRNA sequence into close proximity with the barcode and UMI at the 3′ end. A targeted reverse PCR step is used to re-linearize the circularized molecule in such a way that the 5′ sequence remains near the barcode-UMI junction. This re-linearized molecule is then randomly fragmented, and sequencing adapters are ligated to the resulting fragments to prepare the sequencing library.

Because of this molecular architecture, every fragment, regardless of whether it originated from the 5′ end, middle, or 3′ end of the transcript, retains the original barcode and UMI. During analysis, all reads with matching identifiers are grouped and assembled using a proprietary bioinformatics pipeline, enabling accurate reconstruction of full-length transcripts. This method preserves the integrity of transcript isoforms and allows for high-resolution detection of splice variants, fusion transcripts, and full-length immune receptor sequences using standard short-read sequencing platforms.

Part 3: Comparison of Commercial Kits by Attributes

Number of cells

Above 1,000,000 cells per run:

Leaders: Parse Biosciences, Scale Biosciences, 10X Genomics

Parse Biosciences, Scale Biosciences, and 10x Genomics lead the field in high-throughput single-cell RNA sequencing, each offering kits that process over 1 million cells per run.

Parse's Evercode™ Whole Transcriptome and BCR/TCR Mega series reach up to 1 million cells, with the latest Evercode™ WT PENTA V3, launched in February 2025, extending that capacity to an industry-leading 5 million cells per run.

Scale Biosciences follows closely with its QuantumScale Single Cell RNA Extra Large kit, supporting up to 4 million cells.

10x Genomics’ Chromium Single Cell Flex Gene Expression kit, which employs probe-based chemistry for sensitive transcript detection, supports up to 2.56 million cells per run.

While Parse and Scale leverage combinatorial indexing to scale without instrumentation, 10x remains the only droplet-microfluidics platform to achieve this level of throughput.

Hundreds of Thousands Range

Highlights: BD Rhapsody™ HT Xpress System, 10x Genomics, Takara Bio Shasta Total RNA-Seq Kit, Flexomics, Celldom

In the high 100,000s range, 10x Genomics remains a strong player with its Chromium Single Cell Universal 3′ and 5′ Gene Expression kits. The number of cells per run depends on the sample multiplexing strategy. In singleplex mode, one sample is loaded per channel at an input of approximately 20,000 cells. With 8 channels per chip, this yields around 160,000 cells per chip. In contrast, using sample hashing, where samples are tagged with antibody- or lipid-based barcodes and pooled, each channel can accept up to 60,000 cells, resulting in a total of 480,000 cells and up to 96 samples per chip.

The BD Rhapsody™ HT Xpress System, launched in February 2023, uses microwell-based partitioning and achieves a total throughput of up to 800,000 cells per run. Each of the 8 lanes in the cartridge contains hundreds of thousands of microwells designed to isolate single cells. Microwell-based systems have historically had lower throughput due to higher chip fabrication costs and physical space limitations, but the HT Xpress system overcomes many of these issues with improved cartridge architecture and fluidics control.

Another microwell-based platform is Singleron’s GEXSCOPE® HD Chip, which captures 9,000 to 30,000 cells per sample under standard conditions. With CLindex® multiplexing, a technology that allows up to 16 samples to be pooled into a single chip, the system can support up to 160,000 cells per run. However, cell quality may decline beyond 120,000 cells due to elevated doublet rates (around 28%).

Also operating in this range are smaller kit formats from both Parse Biosciences and Scale Biosciences. Parse’s Evercode™ WT and Mega kits, launched in 2021, support up to 100,000 and 1,000,000 cells per run, respectively. Scale’s Quantum Barcoding Modular Kits, introduced in 2025, enable similar mid-range throughput, with options designed for runs of up to 168,000 or 2,000,000 cells depending on configuration.

In October 2024, Takara Bio launched the Shasta Total RNA-Seq Kit, which supports up to 100,000 cells per run while offering full-length transcript coverage. Unlike traditional 3′ or 5′ kits that capture only transcript ends, Shasta uses SMART-seq chemistry to generate full-length cDNA, which is particularly useful for studies focused on isoform-level expression and transcript structure. Takara Bio’s Shasta Total RNA-seq Kit is likely one of the only full-length mRNA technologies in the market at this throughput.

Finally, in the low 100,000s range, companies like Flexomics (founded in 2019) and Celldom (founded in 2016) offer single-cell technologies that integrate microwell-based isolation with high-content imaging and functional assays. These platforms are purpose-built for specialized applications such as compound screening, immunophenotyping, and mechanistic cell biology, making them valuable tools in translational research.

Thousands to Tens of Thousands Range

In the tens of thousands range, we begin to see more specialized products focused on epigenetic profiling and methylation analysis, often with lower cell throughput due to the complexity and sensitivity of these assays.

10x Genomics offers the Chromium Single Cell ATAC-seq and Chromium Single Cell Multiome ATAC + Gene Expression kits, each capable of profiling up to 80,000 cells per run. These kits use droplet-based partitioning to capture chromatin accessibility and, in the case of Chromium Multiome, gene expression from the same cell.

In contrast, Scale Biosciences provides a Single Cell Methylation Kit that supports up to 16,400 cells per run. Unlike 10x ATAC-seq kits, which detect open chromatin regions via transposase-accessible sites, Scale’s methylation kit uses bisulfite-free labeling of methylated cytosines, enabling direct quantification of methylation without harsh chemical treatment. This gentler, indexing-based workflow preserves DNA integrity and is more compatible with sensitive sample types, though at a trade-off in throughput.

Also within this range is another full-length RNA platform: MobiuSCOPE Full-Length Single Cell RNA Sequencing. Designed to detect complete transcript structures, including 5′ and 3′ ends, MobiuSCOPE supports up to 30,000 cells per run. MobiuSCOPE uses a circularization strategy that enables full-length transcript reconstruction from fragmented reads, an innovative alternative to traditional full-length methods like SMART-seq.

This places Singleron’s MobiuSCOPE it in the same category as Takara’s Shasta platform for isoform-level transcriptomics

Hundreds Range

In the low-throughput category of hundreds of cells per run, BioSkryb Genomics offers its Resolve suite of technologies, including:

• ResolveOME™: Single-cell multiomics kit for simultaneous RNA and DNA sequencing

• ResolveDNA® scDNAseq: High-resolution single-cell whole genome amplification and sequencing

• ResolveDNA® WGA: High-resolution whole genome amplification

BioSkryb's platform relies on a proprietary primary template-directed amplification (PTA) method that uniformly amplifies the entire genome of a single cell with minimal bias. This method enables extremely high coverage and low error rates, making it well-suited for detecting structural variants, rare mutations, and single-nucleotide variants in individual cells.

While most single-cell transcriptomics platforms compete in the hundreds of thousands or millions of cells per run, BioSkryb occupies a niche space focused on high-fidelity DNA and multiomic profiling for dozens to hundreds of cells. Their technology is particularly valuable for cancer genomics, microbial diversity studies, and developmental lineage tracing.

Note: CS Genetics is an exception here and doesn’t quite fit into our comparison. Their SimpleCell technology does not impose a fixed upper limit on the number of cells per experiment. Because the entire process takes place in free solution, without the need for droplet microfluidics or microwells, the number of cells that can be processed is determined primarily by the number of samples a user runs at a time and whether automated liquid handling robotics is used.

Multiomics Support

Single-cell sequencing has evolved beyond basic transcriptomics into a rich multi-omics landscape. Virtually all vendors now support single-cell whole transcriptome RNA sequencing, which remains the cornerstone for profiling cellular gene expression. Many platforms are extending into multi-modal assays, combining gene expression with other analytes like immune receptor sequences, proteins, genomic DNA variants, chromatin state, and even live-cell functional readouts.

Who are the leaders in multi-modal detection?

1. 10X Genomics, Singleron, and BD Biosciences offer the most breadth.

10x Genomics is widely seen as a leader in single-cell multi-omics. Their Chromium platform allows numerous combinations: gene expression (3’ or 5’ RNA-seq) can be paired with CRISPR perturbation readouts (capture of guide RNAs) and with cell-surface protein expression via DNA-barcoded antibodies. 10x also offers a joint assay for chromatin accessibility (ATAC-seq) plus whole transcriptome, enabling epigenomic and transcriptomic profiling in the same cell. Additionally, 10x’s immune profiling kits capture full-length T-cell and B-cell receptor (TCR/BCR) sequences concurrent with mRNA, linking clonal identity to cell state. This breadth makes 10x a benchmark: It supports essentially every analyte in the matrix except DNA methylation and whole-genome DNA sequencing.

Several other vendors have established multi-omic co-detection capabilities as well. BD Biosciences (through the Rhapsody system) enables simultaneous measurement of mRNA, immune receptor repertoire (TCR/BCR), and protein markers in single cells. BD’s system uses a microwell-based capture and has an add-on kit for full-length TCR/BCR profiling that works alongside whole transcriptome or targeted gene expression profiling. Like 10X Genomics, BD also adapted CROP-seq methods to capture CRISPR guide RNAs with mRNA on their platform.

Singleron Biotechnologies has steadily emerged as a major player in integrated single-cell analysis. The company’s platform also supports simultaneous transcriptome profiling with TCR/BCR sequences, surface protein expression, and targeted transcriptome panels, enabling focused gene expression analysis when full-transcriptome depth is unnecessary. A standout feature is its AccuraSCOPE® platform, which supports co-detection of the whole genome and transcriptome from the same cell. This distinguishes it from 10X Genomics and BD Biosciences and positions Singleron as one of the very few vendors to offer such deep genomic-transcriptomic integration.

2. Parse and Scale Biosciences support multi-modal profiling at massive scale (4M+ cells), with Parse offering broader analyte breadth and Scale standing out as the only vendor with single-cell DNA methylation.

Parse Biosciences has emerged with a combinatorial barcoding platform (Evercode) that supports multi-modal readouts at large scale. Parse offers kits for Whole Transcriptome as well as specialized kits for TCR/BCR sequencing (Evercode TCR/BCR, capturing paired receptor chains plus gene expression) They’ve also introduced CRISPR Detect, an add-on allowing single-cell RNA-seq readouts of pooled CRISPR screens by capturing guide RNAs along with transcripts. This puts Parse in the multi-omic arena similar to 10x, albeit using a plate-based barcoding strategy with no custom hardware.

Another new entrant, Scale Biosciences, uses a “Quantum Barcoding” combinatorial index technology. Scale Bio’s platform supports 3’ gene expression, an optional module for CRISPR guide capture (a kit to enrich sgRNAs in parallel with the RNA-seq library), and even a dedicated single-cell DNA methylation sequencing kit–truly the first of its kind on the market. Scale Bio also partnered with BioLegend to enable high-parameter protein detection alongside its RNA-seq: their TotalSeq™ “Phenocyte” solution attaches BioLegend’s DNA-barcoded antibodies to Scale’s workflow to profile dozens of surface proteins per cell concurrently.

3. Mission Bio is the only platform that offers co-detection of targeted genomic DNAs and proteins. Bioskryb provides a robust, new deep “tri-omic” approach (DNA, RNA, and protein profile for every cell).

In terms of genomic DNA and mutation analysis at single-cell level, the clear leader is Mission Bio. While most platforms center on RNA, Mission Bio’s Tapestri platform was uniquely built for targeted DNA sequencing of single cells, with the ability to detect single-nucleotide variants (SNVs), copy number variants (CNVs), and small indels across a panel of genes, for hundreds to thousands of cells. Mission Bio pioneered simultaneous DNA+protein profiling: the platform can amplify target DNA regions of interest and capture protein markers (via antibody-oligo conjugates) from the same cell, linking genotype to immunophenotype.

No transcriptome is measured; this is a DNA-centric multi-omic view, but powerful for applications like cancer, where mutations and surface markers define clones. Mission Bio’s focus on multi-omic genotype + phenotype has influenced others to develop similar “DNA + RNA + protein” approaches.

BioSkryb Genomics has introduced another high-content single-cell multi-omics approach with its ResolveOME kit. This workflow performs unified whole-genome DNA amplification and full-length mRNA cDNA synthesis from the same cell, enabling near-comprehensive sequencing of both the genome and transcriptome of that cell. In practice, a single cell processed with ResolveOME yields a whole-genome sequence (with high coverage uniformity thanks to their PTA technology) together with a full transcriptome (capturing mRNAs end-to-end). BioSkryb further demonstrated adding a third modality: by integrating BioLegend TotalSeq™ antibodies, they can also measure selected surface proteins in those same cells. In other words, BioSkryb’s approach can unite genomic, transcriptomic, and proteomic data from a single cell, a true tri-omic analysis. This is currently a lower-throughput, specialist solution (aimed at deep analysis of tens to hundreds of cells, for example in tumor evolution studies), but it represents a state-of-the-art in multi-omic content per cell.

4. Emerging trend from new entrants (Flexomics and Celldom): Combining live functional cell assays with sequencing.

Finally, new players are combining live-cell assays with sequencing. Celldom and Flexomics both use imaging-based platforms to monitor single-cell behaviors in large arrays of nanowells or picowells, then retrieve those cells for molecular analysis.

Flexomics, for instance, uses a high-density picowell chip with optical barcoding to perform live fluorescence imaging of thousands of single cells (for functional readouts), followed by lysis and capture of mRNA from each well’s cell. This yields paired phenotype and transcriptome data, linking microscopy and sequencing in one workflow. Flexomics emphasizes that any live-cell assay (proliferation, morphological changes, reporter expression, etc.) can be miniaturized on their chip (“Bring Your Own Assay”) and combined with sequencing readouts.

Similarly, Celldom’s platform allows long-term time-lapse imaging of single cells (e.g. to track drug responses or clone formation) in an array of up to millions of nanowells, with an integrated picker to extract individual cells or clones of interest for downstream DNA/RNA sequencing.

These approaches are still emerging, but they offer unique multi-modal data, bridging dynamic cellular function (live-cell phenotypes) with genomic analysis for the first time.

5. Modular multi-omics

Atrandi Biosciences’ Semi-Permeable Capsule (SPC) technology introduces a fundamentally novel paradigm for single-cell multi-omics. Each SPC is a porous hydrogel-based microcapsule that encases individual cells, allowing small molecules and reagents to diffuse in and out while retaining cellular contents. This design enables a sequence of biochemical reactions (lysis, amplification, reverse transcription, bisulfite conversion, and more) to be performed stepwise in the same capsule through simple buffer exchanges. Unlike droplet-based systems that constrain workflows to “one-shot” chemistries, SPCs support modular, multi-step assays under tightly controlled reaction conditions.

While Atrandi Biosciences offers a single-cell DNA microbial and eukaryotic kit, its SPC Innovation Kit allows researchers to mix-and-match and create their own custom, multi-omic single-cell assay.

Example Proof-of-Concepts:

• Whole-genome sequencing: In a 2023 Genome Biology study (Balan et al.), SPCs enabled high-throughput single-cell genome sequencing of over 60,000 bacteria, achieving strain-resolved assemblies with minimal cross-contamination.

• Phenotype-genotype linkage: A Nature Microbiology study (Orakov et al.) demonstrated culturing microbial cells within SPCs, monitoring their growth phenotypes, and linking those to genomic identity.

• Targeted gene expression (RNA cytometry): In a 2023 Nucleic Acids Research study (Leonaviciene & Mazutis), SPCs enabled in-capsule RT-PCR with fluorescent primers for digital detection of specific transcripts.

• Protein detection: Atrandi’s white papers confirm theoretical compatibility with antibody-based staining, including DNA-barcoded antibodies, supporting workflows similar to CITE-seq.

• Targeted genomic assays: SPCs can also be used for multiplexed genotyping via barcoded PCR, hybrid capture, and concatenated sequencing strategies.

• Live-Cell Assays: In a Lab on a Chip 2020 study, bacterial clones were cultured and phenotyped inside SPCs before sequencing, demonstrating the ability to maintain viability and assay functionally.

• Atrandi emphasizes SPCs as miniature bioreactors, capable of supporting high-throughput single-cell cultivation, observation (e.g. fluorescence or microscopy), and downstream analysis.

Software and Bioinformatics Support

Almost every vendor now offers its own in-house single-cell analysis platform to help customers realize the value of their kits more quickly.

That said, in many cases, it feels like each provider is re-inventing the wheel, primarily to gain the benefits of white-labeling and full control over how their analyses are presented. A great deal of bioinformatics talent ends up being spent on software engineering, UI development, and cloud hosting, tasks that should be easy to standardize but often become significant time sinks, pulling focus away from core computational biology or assay development work.

While platforms like BioTuring and ROSALIND exist, they are generic third-party tools. Using them often means sacrificing flexibility in algorithm design or visualization customization, both of which could be critical for showcasing the unique features of a vendor's kit.

There is room for innovation here, not just to build better visualization layers, but actual frameworks. Imagine developer-first products that provide clean APIs and tooling to deploy analysis pipelines quickly. This idea is not new in the tech sphere. Platforms like Vercel or Streamlit have done it well for abstracting a way the cloud hosting and UIs construction with simple APIs. Still, there is nothing purpose-built for large-scale single-cell analysis. Features like fast HDF5 operations, interactive annotation in a browser-based viewer, or lasso-selecting cells within a UMAP all remain fragmented across custom implementations.

A well-designed framework could dramatically reduce the time teams spend re-building infrastructure and let them focus instead on real science and (cool) impactful algorithms. (Note: this is a topic I spend a lot of time thinking about. If it resonates, reach out!)

Parting Thoughts

The landscape is getting crowded, but that's a good thing. It reflects a field moving fast, driven by real biological questions and the growing demand for resolution, scale, and functional depth. Each kit listed here represents a specific view on what matters most in single-cell biology: throughput, analyte diversity, multiomics, FFPE compatibility, software support, or live-cell function. No single platform will be best for every use case. We hope that through this article and the single-cell landscape database, researchers can more easily navigate the expanding ecosystem of technologies. Our goal is to help scientists find the right tools for their questions and to recognize the builders, often working quietly behind the scenes, who make those tools possible.

However, accurately classifying cells is a complex task. Gene expression levels are continuous rather than discrete, following gradient changes. Technical variations in experiments can introduce transcriptional differences that lack biological relevance [2].

The traditional approach relies on annotating cell types based on marker gene expression, but this method can yield conflicting results depending on tissue type, disease state, and experimental design. As a result, researchers must conduct extensive literature reviews and perform significant manual labor to discern real cell types [3]. ML-based approaches have risen in popularity, but differences in reference databases used for model training and ML algorithms can lead to varying annotations, making manual curation unavoidable.

popV: A ML Ensemble for Cell Type annotation

Popular Vote (popV) [4] is a new ML-based tool developed by the Yosef Lab at the Weizmann Institute of Science to address the limitations mentioned above. It employs an ensemble of prediction models with an ontology-based voting scheme to label cells, highlight areas of uncertainty that may require manual scrutiny, and reduce the burden of manual inspection.

Overview of popV’s eight models

popV employs eight machine learning models to transfer labels from an annotated reference dataset, generating predictions alongside a consensus score that reflects confidence in the results. The authors of popV suggest that disagreement across methods often indicates an inaccurate annotation, while agreement among methods typically signals a correct cell type assignment. [4].

The ensemble includes both classical and deep learning-based classifiers, as well as methods that incorporate integration and transfer learning. For example, scANVI is a deep generative model that performs probabilistic label transfer while accounting for uncertainty [5], and OnClass uses ontology-aware classification to assign labels even to unseen cell types [6]. Celltypist is a lightweight logistic regression classifier optimized for speed and interpretability [7]. The remaining classifiers such as SVM [8] and XGBoost [9] are classical supervised learning algorithms known for their strong performance on high-dimensional datasets.

The tool offers three modes for label transfer. Retrain mode trains all classifiers from scratch. Inference mode uses pretrained classifiers to annotate both query and reference cells while constructing a joint embedding with all integration methods. Fast mode utilizes only methods with pretrained classifiers to annotate query cells.

Ground truth data: 718 Hand-Annotated PBMCs Samples from CS Genetics

Peripheral blood mononuclear cells (PBMCs) are widely regarded as the canonical dataset for single-cell RNA-seq analysis. They are easily accessible and relevant to a wide range of scientific questions, including immune responses, disease states, and therapeutic effects. Because of their well-characterized cell type diversity, PBMCs have become the standard for validating new single-cell technologies. In this work, we use PBMC data generated with the CS Genetics SimpleCell™ 3’ Gene Expression assay to evaluate the performance of popV.

718 PBMC samples, processed as 26 experiments, were collected from 16 donors and processed through the CS Genetics SimpleCell™ 3’ Gene Expression assay. The datasets were clustered on a per-experiment basis using Seurat with each cluster manually annotated with a cell type by cell biologists using marker gene expression with individual cells inheriting the annotation of their cluster. This resulted in 24 unique cell types, including a classification of “multiplets” which are assigned when cells had signatures for multiple cell types (5.7% of cells). These manual labels serve as the ground truth for benchmarking model performance. After processing all 26 experiments in Seurat, all cells were combined into a single count matrix of 1,689,880 cells covering 28,340 unique genes.

Cell type labels were mapped to their corresponding Cell Ontology by searching for the manual annotation label using the EMBL-EMI Cell Ontology Lookup Service and selecting the closest match (taking into account relevant markers where appropriate).

Evaluating popV Performance

Here, we compare popV accuracy to cell-biologist-derived annotations of CS Genetics PBMC scRNA-seq data by performing the following analysis:

• Compare pool-based vs. experiment-based training and testing split strategies.

• Validate the OnClass model and its ability to propagate unseen cell labels. We chose to highlight the OnClass model specifically because it has the ability to predict cell types that doesn’t exist in the reference, a common challenge for single-cell datasets.

• Assess label confidence when query datasets contained cell types absent from the reference dataset.

• Investigate the impact of training dataset size, run mode on model performance.

To evaluate model performance, we calculated accuracies1, weighted accuracies2, and stratified accuracies3 using two different majority voting systems (simple majority voting, popV consensus scoring) and three run modes (retrain, inference, fast).

Pool-based vs. experiment-based splits

To evaluate the performance of the ensemble-based annotation approach, we trained models on subsets of the manually annotated PBMC dataset and assessed their ability to predict cell types in held-out cells. Specifically, we trained models using either 10% (126,435 cells) or 50% (632,175 cells) of the reference dataset and predicted cell types for 39,987 query cells.

To better understand generalizability, we compare two training/testing split strategies while maintaining a similar number of total cells in each setup:

• Pool-based splitting: A random sample of 50% of all cells was taken and split into 80% training and 20% testing.

• Experiment-based splitting: 20 (1,264,350 cells) out of 26 experiments were randomly selected for training, with the remaining 6 experiments (425,530 cells) used as unseen query data.

The hypothesis is that experiment-level splitting better simulates true model performance on unseen data and that the accuracy observed with the pool-based approach may be inflated due to test cells coming from the same experiments as the training data. However, similar accuracies were observed across both approaches.

Initial popV Annotation Results

The following results are based on the experiment-level split, using a subset of 39,987 query cells and a training dataset of 126,435 cells.

For well-characterized cell types such as classical monocytes, memory B cells, and CD8-positive alpha-beta memory T cells, popV’s ensemble models reached high consensus scores, with nearly all eight models agreeing on their labels.

In contrast, cells located between similar clusters, such as CD8+ αβ cytotoxic and memory T cells, showed low consensus among models, reflecting a challenge in manual annotations that rely on cluster-level markers (Fig. 1). These boundary cells often don’t align cleanly with manual cluster labels, making direct comparisons less meaningful.

This highlights the advantage of cell-level annotation, where each cell is labeled individually rather than assigning a single label to an entire cluster. This approach can yield more accurate results, especially for cells that lie at the boundary between clusters and may express mixed marker profiles.

Ontology-Aware Predictions with OnClass

OnClass is one of the eight models available in popV. It stands out from other models by requiring all reference labels to correspond to valid terms in the Cell Ontology. This promotes standardized, ontology-compliant annotations, a critical advantage for data harmonization across studies. However, this requirement can pose challenges when working with ambiguous or custom labels. In our CS Genetics dataset, some cells were manually annotated as “multiplets” or “proliferating cells”, terms that don’t have corresponding entries in the Cell Ontology and therefore cannot be used during OnClass training.

Despite this challenge, a key strength of OnClass is its ability to predict cell types absent from the reference dataset. It first assigns an annotation from the reference labels (referred to as OnClass_seen) before propagating it to identify a potentially more refined label in the Cell Ontology (even if this label is absent from the reference; referred to as OnClass_prediction), thus generating two different annotations.

When popV includes OnClass in its ensemble, there are two weighting approaches to determine the final annotated labels.

1. Simple majority voting: All model predictions are weighted equally, and only the OnClass_seen label is used.

2. popV consensus scoring: OnClass exerts more influence by contributing both OnClass_seen and OnClass_prediction, with additional weight given based on the path through the ontology hierarchy.

We proceeded to evaluate OnClass performance to decide whether to include OnClass as one of the models contributing to popV's consensus.

How accurate is OnClass model prediction, and how well can it propagate unseen cell labels?

Assessment of OnClass model prediction

We assessed the accuracy of both the OnClass_seen prediction (i.e. a direct comparison to the manual annotations) and OnClass’s ability to accurately assign cell types through propagation (the OnClass_prediction annotation).

We first evaluated the accuracy of OnClass_seen prediction performance against the ground truth (Fig. 2a). To do this, we ran popV inclusive of the OnClass model, using a 10% downsampling of both the 20 experiment reference and 6 experiment query sets. The OnClass_seen annotations showed a 77% accuracy compared to manual annotations.

Assessment of OnClass ability to propagate unseen cell labels

To evaluate OnClass's ability to propagate cell type labels along the ontology hierarchy, we retrained the model using a modified reference dataset. Instead of assigning each cell to its original, fine-grained ontology node, we reassigned each one to its parent node in the ontology tree. This allowed us to test whether OnClass could correctly predict the original, more specific cell type from the broader parent label.

In general, OnClass performed poorly at recovering the original, granular cell types. For example:

• Most T cells were labeled as immature T cells

• Monocytes were labeled as intermediate monocytes

• Mature B cells were often mislabeled as germinal center B cells (Fig. 2b)

OnClass struggled to identify finer subtypes like memory T cells or naive thymus-derived αβ T cells. There were also clear misclassifications:

• CD8+ αβ cytotoxic T cells were mislabeled as regulatory T cells

• Natural killer cells were classified as innate lymphoid cells

• Platelets were incorrectly labeled as hematopoietic cells, classical monocytes, or even immature T cells (Fig. 2c)

These results underscore the limitations of OnClass’s propagation strategy, particularly in accurately predicting specific or unseen cell types, supporting the authors’ concerns about its reliability in high-resolution annotation tasks.

The Verdict

Despite the poor performance of the propagation approach, OnClass was included as one of the models in popV, given that its OnClass_seen predictions still yielded promising results (Fig. 2d). Consequently, we adopted simple majority voting, rather than consensus scores that also incorporate OnClass_predicted, as the weighting strategy and default setting for cell labeling in the bioinformatics workflow.

How does PopV handle ambiguous cell types?

As noted earlier, CS Genetics biologists’ original annotations include ambiguous labels such as "multiplets" and "proliferating cells” when a cluster contains signature markers from multiple different cell types.

To test whether ambiguous labels in the query dataset, such as “multiplets” and “proliferating cells”, would lead to inaccurate annotations, we trained OnClass on a reference dataset that excluded these labels. We then examined how the model annotated the ambiguous cells in the query.

We found that popV assigned a wide range of cell types to these ambiguous cells (Fig. 3), and these assignments were consistently associated with low consensus scores. This low-confidence behavior is valuable: unlike the high-confidence assignments seen for well-defined cell types (Fig. 3), it serves as a built-in signal for users to flag uncertain or mixed-cell cases for further review or exclusion from downstream analysis.

Optimizing popV Parameters

Consensus and Majority Voting yield comparable accuracy

popV supports three modes: retrain, inference, and fast mode, depending on model availability.

In inference mode, we compared the performance of the consensus score and majority voting for cell type annotation:

• With the popV consensus score:

  • Accuracy in the Very High confidence bin was ~93%

  • ~60% of cells fell into this bin (Fig. 5a)

• With majority voting:

  • Accuracy was ~92%

  • ~69% of cells fell into this bin (Fig. 5b)

These results are consistent with previous reports from the popV authors, who observed over 95% accuracy when 7 or 8 out of 8 models agreed—criteria that correspond to our Very High confidence bin.

All accuracy values were calculated using 39,987 query cells, excluding ambiguous labels like proliferating cells and multiplets. When including these ambiguous cells (totaling 42,544 cells):

• Consensus score accuracy dropped slightly to 90%, with 59% of cells in the Very High confidence bin

• Majority voting had 89% accuracy, with 67% of cells in this bin

Expanding training set size doesn’t improve performance

No significant differences in accuracy were observed between models trained on 10% (126,435) and 50% (632,175) of the reference dataset despite a considerable increase in compute time and cost (Table. 1).

This non-significant result may be due to the parameter N_samples_per_label, which controls the number of sampled cells per label, and is set to 500 by default, limiting the impact of increasing the reference pool size. To test this, we increased it to 2500. However, no significant differences were observed. Thus a more likely explanation for the lack of difference between the 10 and 50% training sets is that all cells have relatively similar gene expression profiles, so that increasing the number of cells used for training contributes minimally to performance improvements.

Fast mode offers cost-efficient annotation with minimal accuracy trade-off

Finally, we compared the performance of inference and fast modes in terms of accuracy, computational efficiency, and cost. The inference mode consistently achieved slightly higher accuracy than the fast mode, which is expected given that the fast mode applies a single training epoch in scArches for cell annotation. However, this improvement in accuracy was marginal, with only a ~3% increase when annotating 39,987 cells using 10% of the reference data. In contrast, the inference mode required nearly three times the computational time compared to the fast mode, resulting in significantly higher resource consumption. These findings highlight a trade-off between accuracy and computational efficiency, where the fast mode offers a more cost-effective alternative with only a minimal reduction in accuracy.

a.

b.

The table below summarizes the accuracies, run time and cost between inference and fast mode with 1 L40s GPU. 1 credit is equivalent to $1.

To Conclude

This case study demonstrates that popV confidently annotates most cells accurately while offering mechanisms to identify ambiguous cells requiring further consideration. Given the complexity of cell annotation and the associated expert labor burden, popV, which leverages ensemble ML models and ontology-based consensus voting, can help reduce the manual annotation workload. For cell types that are not present in the training dataset, correctly labeling the cells remains a challenge, but popV mitigates false positives by flagging uncertain cases.

Beyond general annotation, automatic cell typing tools like popV have practical utility in areas like drug discovery, where accurately identifying disease-relevant cell populations and tracking subtle transcriptional changes is crucial for target validation and therapeutic development. Also, in kit and assay validation, popV offers a scalable way to benchmark single-cell assays. By comparing predicted cell types distribution across batches, labs can quickly detect deviations in sample quality.

All eight pretrained models trained on CS Genetics PBMC data, along with the popV workflow, are available on latch.bio, meaning you can start annotating your PBMC data directly on LatchBio.

Get started with popV: Annotate your PBMC single-cell data now on the LatchBio platform. Click here.

For more information on CS Genetics’ SimpleCell technology, please visit csgenetics.com.

Thank you to Imogen West, Ben Hume and Mike Stubbington, who curated CS Genetics PBMC dataset and provided thoughtful review.

Reference

1: https://pubmed.ncbi.nlm.nih.gov/29674595/

2: https://www.csbj.org/article/S2001-0370(21)00019-2/fulltext

3: https://www.sciencedirect.com/science/article/pii/S2001037021000192

4: https://www.nature.com/articles/s41588-024-01993-3

5: https://www.nature.com/articles/s41592-018-0229-2

6: https://pubmed.ncbi.nlm.nih.gov/34548483/

7: https://pmc.ncbi.nlm.nih.gov/articles/PMC7612735/

8: https://dl.acm.org/doi/10.5555/1953048.2078195

9: https://dl.acm.org/doi/10.1145/2939672.2939785

Recent technologies have transformed our ability to visualize protein expression and modification directly within intact tissue. What was once an abstract proteome map now becomes a richly detailed tissue blueprint revealing protein-protein interactions, signaling networks, and cellular neighborhoods.

This article is the third installment in a series investigating different “branches” of spatial technology. In part 1 and part 2 of the series, we focused on the history and commercialization of sequencing-based and imaging-based spatial RNA technologies.

In this post, we turn our attention to spatial proteomics: How do we see proteins in their native tissue context?

To help readers navigate the spatial landscape as a whole, we’ve extended our website spatialbiologylandscape.com of sequencing and imaging transcriptomics technologies to include the 14 proteomics products describe here.

This article will be organized by core concepts, methods by history and methods by attributes.

Please see this table of content to see the overall structure of the article and quickly jump to the part you’re interested in:

1. Technology Tree

   1. Fluorescence Imaging-based Methods

   2. Sequencing-based Methods

   3. Mass spectrometry-based Methods

2. History and Key Technologies

   1. Fluorescence Imaging-based Methods

      1. Fluorophore Tags Technologies

         1. Traditional Immunohistochemistry

         2. Multiplexing IHC

         3. MELC (Multi-Epitope Ligand Cartography) (2006)

         4. MxIF (Multiplexed Immunofluorescence) (2013)

         5. t-CycIF (Tissue-based Cyclic Immunofluorescence) (First published in 2013, refined in 2018)

         6. sIHC (Sequential Immunohistochemistry)

         7. SWITCH (System-Wide control of Interaction Time and kinetics of CHemicals) - Introduced in 2015

         8. Array-Tomo - Introduced in 2018

         9. 4i (Iterative Indirect Immunofluorescence Imaging) - Published in 2018

         10. IBEX (Iterative Bleaching Extends Multiplexity) - Introduced in 2022

      2. DNA Tags Technologies

         1. CODEX (2018)

         2. ImmunoSABER (2019)

         3. DEI (DNA Exchange Imaging)

         4. HCR (Hybridization Chain Reaction) - First adapted for protein detection around 2017

         5. in situ Hybridization Chain Reaction (isHCR) (2020)

   2. Sequencing-based Methods

      1. Spatial CITE-seq

      2. Molecular Pixelation

      3. DSP

   3. Mass Spectrometry-based Methods

      1. Label-Free Mass Spectrometry Approaches

         1. LC-MS/MS

            1. LCM-SISPROT

            2. MASP: Micro-Scaffold Assisted Spatial Proteomics

            3. scDVP: Single-Cell Deep Visual Proteomics

         2. Imaging Mass Spectrometry (IMS)

            1. Top-down IMS

            2. Bottom-up IMS

      2. Metal-Tagged Mass Spectrometry Approaches

3. Comparison between Commercial Solutions

   1. Number of Targets & Throughput

   2. Panel Flexibility and Customization for Antibodies

   3. Tissue Types

   4. Maximum Scan Area per Run

   5. Cost

   6. Multiomic Capabilities

4. Future Perspectives

Technology Tree

To navigate the spatial proteomics landscape, we divide it into three main categories:

1. Fluorescence Imaging-based

2. Sequencing-based

3. Mass spectrometry-based.

Fluorescence Imaging-based Methods

These approaches rely on protein detection via antibody labeling coupled with fluorescent microscopy. Traditional imaging techniques are limited to detecting only 5-7 proteins simultaneously due to the spectral overlap constraints of conventional fluorophores. To overcome this limitation, multiplexing strategies have emerged, following a cyclical workflow: tissue samples are stained with antibody-fluorophore conjugates, imaged to capture fluorescent signals, then subjected to antibody removal or fluorophore deactivation before repeating the process through multiple rounds to expand protein detection capacity.

Sequencing-based Methods

Sequencing-based methods involve the use of next-generation sequencing machines to quantify proteins. Unlike imaging-based methods requiring multiple staining cycles, sequencing-based techniques typically require only a single tissue staining step using specialized antibodies. Each antibody carries a unique DNA barcode that corresponds to its target protein. These DNA sequences can be amplified in situ and subsequently read out via NGS platforms.

Mass spectrometry-based Methods

Mass spectrometry (MS) approaches to spatial proteomics operate on a fundamental principle: Proteins can be identified and quantified by measuring their mass-to-charge ratios.

• The process typically begins with a tissue section that undergoes controlled protein extraction or direct ionization from specific locations.

• The extracted proteins are converted to gas-phase ions, which are then separated based on their mass-to-charge ratios using electric or magnetic fields. The pattern of separated ions creates a "molecular fingerprint" that can be matched to known protein signatures.

By systematically analyzing multiple positions across the tissue, researchers can construct detailed maps showing which proteins appear where, revealing crucial information about cellular organization and function without requiring antibody labeling. This approach offers particular advantages for discovery-based research, as it can detect proteins without prior knowledge of which ones might be present in the sample.

History and Key Technologies

Now that we understand how each category of method works at a high level, we will dive into the history and the evolution of key technologies under each branch.

1. Fluorescence Imaging-based Methods

1.a. Fluorophore Tags Technologies

Traditional Immunohistochemistry

Before diving into the latest advances in multiplexed protein detection, it is worthwhile to revisit the basic principles of traditional immunohistochemistry (IHC).

Immunohistochemistry (IHC) was first developed in the 1940s by Albert H. Coons and colleagues. In 1941, Coons successfully labeled antibodies with fluorescent dyes, allowing the visualization of antigens in tissue sections. This was the foundation of immunofluorescence, which later evolved into modern IHC techniques using enzyme-linked antibodies (such as peroxidase or alkaline phosphatase) for chromogenic detection in the 1960s and 1970s.

Immunohistochemistry (IHC) follows these key principles:

First, a fixed tissue sample is exposed to antibodies that target specific antigens. This typically involves a primary antibody that binds to the target, followed by a secondary antibody carrying either a fluorescent tag (visible under fluorescence microscopy) or an enzyme. With enzyme-labeled antibodies, adding the appropriate substrate creates a colored deposit exactly where antibody binding occurred.

For some applications, directly labeled primary antibodies can be used, skipping the secondary antibody step, though this often results in weaker signals.

Multiplexing IHC

Despite its utility, traditional IHC faces several technical hurdles:

• Spectral overlap between fluorophores, limiting the number of detectable targets

• Cross-reactivity of antibodies, leading to non-specific staining

• Limited multiplexing capacity, restricting simultaneous detection of multiple markers

• Tissue autofluorescence, reducing signal clarity in fluorescence-based IHC

These problems mean only 3-5 proteins can be detected in an IHC experiment.

To solve this, the concept of multiplexing IHC emerged: Instead of staining for just one or two markers at a time, this method uses multiple cycles of staining, imaging, and washing. In each cycle, specific antibodies bind to their target proteins and are visualized using fluorescent or chromogenic labels. After imaging, the markers are removed, and the process is repeated with new antibodies. By layering these cycles, researchers can build a detailed picture of how different proteins are distributed and interact within the tissue.

Over the past decades, many strategies for multiplexed IHC have been developed, such as MELC, MxIF, t-CycIF, sIHC, SWITCH, Array-Tomo, 4i, and IBEX. We will walk through the key concepts of how each works below.

MELC (Multi-Epitope Ligand Cartography) (2006)

One of the first examples of multiplexed IHC is Multi-epitope ligand cartography (MELC), a fully automated robotic platform that enables highly standardized iterative immunofluorescence imaging.

The process begins with a fixed tissue sample where a fluorophore-labeled antibody is applied, imaged at precise coordinates, then completely photobleached to eliminate signal before the next antibody cycle. This automation enables detection of 100+ proteins on a single tissue section with perfect spatial registration between images.

The data processing pipeline integrates these sequential images into a multidimensional dataset where each pixel contains information about dozens of proteins present at that precise spatial location. Downstream computational analysis then identifies protein expression patterns and co-localization networks that reveal functional cellular neighborhoods.

Despite its powerful capabilities, MELC does face several limitations. The sequential staining process is time-consuming, often requiring multiple days for high-plex imaging. Photobleaching, while gentler than chemical inactivation, can still cause cumulative tissue damage over many cycles, potentially affecting antibody binding in later rounds. There is also the challenge of antibody validation across many targets, as cross-reactivity becomes increasingly problematic with higher target numbers. The extensive automation, while reducing human error, requires specialized equipment not widely available in standard laboratories.

MxIF (Multiplexed Immunofluorescence) (2013)

Unlike MELC, which uses photobleaching for signal removal, MxIF uses alkaline oxidation chemistry for fluorophore inactivation and requires background autofluorescence imaging before each staining round. Although MxIF’s chemical approach for fluorophore quenching offers more complete signal elimination, it may also cause greater cumulative tissue damage compared to MELC's light-based photobleaching.

After imaging, the fluorescent dye is inactivated, creating a clean slate for the next antibody set. This approach has demonstrated the capability of analyzing up to 61 different proteins in single tissue samples and has been successfully applied to reveal heterogeneity in ductal carcinoma breast cancer samples.

t-CycIF (Tissue-based Cyclic Immunofluorescence) (First published in 2013, refined in 2018)

t-CycIF represents a major shift toward accessibility and scalability in multiplexed IHC. Unlike MELC and MxIF, which rely on specialized automation and proprietary chemistry, t-CycIF uses standard reagents, commercially available antibodies, and open-source software, making it significantly more cost-effective and widely applicable.

The process follows a cyclic workflow similar to MELC and MxIF: staining with fluorophore-conjugated antibodies and imaging, followed by fluorophore inactivation by oxidation (with hydrogen peroxide), alkaline pH, and light, and a final wash step. Subsequent rounds of staining, imaging, and removal are then performed. This is usually done with four-channel imaging, where 3 channels are used for different antibody-conjugated fluorophores, and the 4th channel is used for a counterstain, such as Hoechst.

Unlike MELC and MxIF, which focus on fixed tissues, CycIF can also be applied to live-cell studies and has been used to study the immune microenvironment of brain tumors and SARS-CoV-2 viral infection in macaque lungs. An extension of CycIF was published in 2018 that is applicable to FFPE tissues, called t-CycIF. The use of standard reagents and commercially available antibodies make CycIF more accessible and applicable to a wide variety of contexts.

sIHC (Sequential Immunohistochemistry)

Sequential Immunohistochemistry (sIHC) is a multiplexed protein detection technique that extends traditional IHC workflows by using iterative cycles of antibody staining, detection, imaging, and antibody removal. Unlike fluorescence-based cyclic methods such as MELC or t-CycIF, which rely on fluorophore bleaching or chemical inactivation, sIHC preserves tissue integrity by using gentle antibody elution buffers between cycles. This allows for the progressive accumulation of protein expression data while ensuring that tissue architecture remains intact. Notable atlases constructed based on principles of sIHC include Cancer Tissue Atlas developed by Gerdes et al. (2013), The Human Protein Atlas, or head and neck squamous cell carcinoma atlas by Tsujikawa et al. (2017).

One of sIHC’s key advantages is its clinical applicability. Because it uses standard FFPE (formalin-fixed paraffin-embedded) samples and widely available reagents, it can be easily integrated into diagnostic workflows without requiring specialized equipment. This has made sIHC particularly valuable in cancer pathology, where analyzing the spatial relationships between tumor cells, immune infiltrates, and stromal components can provide critical insights for treatment decisions. By bridging the gap between research and routine diagnostics, sIHC has helped bring multiplexed protein imaging into real-world clinical practice.

SWITCH (System-Wide control of Interaction Time and kinetics of CHemicals) - Introduced in 2015

While previous techniques enable multiplexed protein detection in thin tissue sections, it is inherently limited by antibody penetration depth—a challenge when working with thicker samples or whole organs. SWITCH, introduced in 2015, overcomes this limitation by introducing precise temporal control over molecular interactions, allowing for uniform immunolabeling of intact three-dimensional tissues.

SWITCH operates through a unique ON/OFF buffer system that can dynamically regulate antibody-tissue interactions:

• OFF state: A global chemical environment prevents unwanted antibody binding, allowing for sample preparation and clearing without premature staining.

• ON state: The system is switched to allow controlled antibody binding, ensuring deep and even penetration across thick tissue volumes.

This method enables whole-organ immunolabeling with subcellular resolution, a breakthrough for spatial proteomics and large-scale tissue mapping. Unlike traditional immunostaining methods that suffer from uneven labeling—especially in deeper tissue regions—SWITCH allows researchers to achieve uniform staining across entire tissue structures while preserving endogenous molecules and tissue architecture.

In the original publication by Murray et al. (2015) in Cell, researchers demonstrated SWITCH's capabilities by imaging more than a dozen proteins across intact mouse brains, revealing brain-wide cellular networks and their molecular characteristics. Since then, the technique has been applied to mapping neural circuits, tracking disease-related protein aggregates throughout brain volumes, and analyzing the three-dimensional organization of diverse tissue types.

Array-Tomo - Introduced in 2018

Another approach for 3D protein imaging is Array Tomography (ArrayTomo), which combines physical sectioning with iterative immunofluorescence. The method begins with embedding tissue in resin and cutting it into ultrathin sections (50-200 nm), which are then mounted as ordered arrays on glass slides. These ultrathin sections overcome the axial resolution limitations of conventional light microscopy and reduce the problem of antibody penetration encountered in thicker tissue preparations. Each section undergoes multiple rounds of antibody staining, imaging, and antibody elution using a mild solvent that strips antibodies while preserving tissue structure and protein epitopes.

This strategy enables reconstruction of three-dimensional protein distributions at subcellular resolution, providing insights into protein localization that would be impossible with single-plane imaging techniques. The method allows for the imaging of dozens of proteins across the same tissue volume while maintaining precise spatial registration. Additionally, the stable preservation of the tissue in resin permits indefinite storage and revisiting of samples for additional markers long after initial processing.

4i (Iterative Indirect Immunofluorescence Imaging) - Published in 2018

While sIHC and t-CycIF advanced multiplexed immunostaining by enabling sequential antibody cycling and signal removal, they still faced trade-offs between signal intensity and antibody availability. sIHC relies on chromogenic detection, limiting multiplexing capacity, while t-CycIF uses direct fluorophore-conjugated antibodies, which can sometimes yield weaker signals. Unlike earlier methods that compromise between signal strength and multiplexing capacity, 4i employs indirect immunofluorescence, where unlabeled primary antibodies bind targets before fluorophore-conjugated secondary antibodies attach, creating significant signal amplification that reveals even low-abundance proteins with exceptional clarity.

The true innovation of 4i lies in its discovery and prevention of photocrosslinking, a previously unrecognized phenomenon where fluorophore excitation generates oxygen radicals that covalently bind antibodies to tissues. By developing specialized imaging buffers containing radical scavengers and crosslinking acceptors, 4i enables complete antibody removal between cycles without harsh chemicals that degrade samples, preserving epitope integrity across at least 20 iterations.

This technological advancement allows 4i to simultaneously capture information across multiple biological scales—from population-level cell crowding effects to subcellular protein organization with nanometer precision—all in the same experiment. The method uses standard off-the-shelf antibodies without special conjugations, making it highly accessible and cost-effective for research laboratories.

In their landmark Science publication, Gut et al. demonstrated 4i's ability to map 40+ proteins in thousands of cells with perfect spatial registration. Their computational approach generates Multiplexed Protein Maps (MPMs) that quantify protein organization at subcellular resolution, revealing previously hidden relationships between protein localization patterns and functional cellular states. This has proven particularly valuable for understanding complex processes like signal transduction pathways, nuclear organization dynamics, and how cells respond to environmental changes or drug treatments.

IBEX (Iterative Bleaching Extends Multiplexity) - Introduced in 2022

Iterative Bleaching Extends Multiplexity (IBEX) is an innovative technique for high-content imaging that enables visualization of over 65 parameters in diverse tissue types. Developed as an accessible alternative to proprietary systems, IBEX employs chemical bleaching with lithium borohydride (LiBH₄) to inactivate fluorophores between imaging cycles.

The method follows a straightforward workflow of iterative antibody labeling, imaging, and chemical bleaching that can be completed in 2-5 days by researchers with basic laboratory skills. IBEX is compatible with over 250 commercially available antibodies and 16 unique fluorophores, offering remarkable flexibility for comprehensive tissue analysis.

A key advantage of IBEX is its open and flexible nature, allowing implementation on various imaging platforms using standard slides and non-proprietary imaging chambers. The protocol can be executed either manually or through an automated workflow using cost-effective instruments and reagents.

IBEX has been successfully applied to characterize complex tissue architectures and cellular ecosystems in both human and animal tissues, supporting initiatives like the Human Cell Atlas while providing valuable insights into tumor microenvironments and tissue histopathology.

The developers provide detailed protocols, validated antibody panels, and open-source image registration software to facilitate widespread adoption of this technology across academic and clinical research settings.

1.b. DNA Tags Technologies

While fluorescence-based protein detection methods have revolutionized tissue imaging, they face inherent limitations in multiplexing capacity due to spectral overlap between fluorophores. This constraint led to the development of DNA-tagging technologies that leverage the vast combinatorial potential of nucleic acid sequences rather than limited fluorescence channels.

In DNA-based methods, tissue preparation involves just a single staining step where antibodies conjugated with unique DNA oligonucleotide tags are applied to the sample.

• After this initial staining, the detection process uses multiple rounds of imaging without requiring additional antibody applications.

• During each imaging cycle, fluorescently labeled reporter probes that are complementary to specific DNA tags hybridize to their target sequences, generating detectable signals.

• When these hybridization events occur, the resulting fluorescent patterns serve as unique identifiers for different proteins within the tissue.

• Between imaging cycles, the reporter probes are removed through rapid chemical or enzymatic processes, allowing new probes to bind in subsequent rounds.

These technologies offer superior multiplexing capability (40+ parameters in a single experiment) and reduced photobleaching concerns compared to traditional fluorescence approaches. However, they typically require specialized instruments, proprietary reagents, and custom antibody conjugation, increasing technical complexity and cost.

Examples of DNA-based technologies include CODEX, ImmunoSABER, and HCR.

CODEX (2018)

CODEX represents a different strategy to multiplexed protein imaging that uses DNA-barcoded antibodies and microfluidics for spatial protein profiling.

Unlike methods that require repeated antibody staining, CODEX performs a single initial staining step where all target proteins are labeled at once using antibodies conjugated to unique DNA barcodes. These barcoded antibodies do not carry fluorophores, making the tissue ready for iterative cycles of imaging without additional antibody application.

Using a microfluidics system CODEX is able to automate the introduction and removal of fluorescent DNA "reporter" oligonucleotides. Each cycle follows these steps:

1. A set of three fluorescently labeled DNA reporters, complementary to three specific antibody barcodes, is flowed into the sample via a microfluidic system. These reporters temporarily bind to their matching barcodes, making the corresponding protein targets fluorescent and visible under a microscope.

2. The sample is imaged, capturing the spatial distribution of the three selected protein markers.

3. A cleaving buffer (chaotropic solvent) is then introduced through the microfluidics system to strip away the fluorescent reporters, effectively removing the fluorescence signal while leaving the DNA-barcoded antibodies intact.

4. A new set of fluorescent reporters is flowed in, binding to a different set of three barcoded antibodies, and the cycle repeats.

This iterative process continues until all markers in the antibody panel have been imaged. The final dataset is reconstructed by computationally aligning and merging the images from all cycles, creating a high-dimensional spatial protein map.

CODEX typically allows 40+ protein markers in one tissue, limited mainly by antibody availability. The technology has since been commercialized and expanded to detect 100+ proteins in a single tissue section while maintaining subcellular resolution.

ImmunoSABER (2019)

ImmunoSABER combines DNA-barcoded antibodies with programmable signal amplification to enhance detection sensitivity for low-abundance proteins, resulting in 10–100x signal amplification in comparison to traditional immunofluorescence techniques.

“SABER” stands for “Signal Amplification By Exchange Reaction”. In immuno-SABER, antibodies are conjugated with short DNA oligonucleotide handles. After binding these antibodies to the tissue, a DNA polymerization reaction is used to grow a concatemer (a long repeated DNA sequence) on each oligo handle. This concatemer has many repeated binding sites. Then, complementary fluorescently labeled oligonucleotide probes (imagers) are hybridized to the concatemers, resulting in amplified fluorescence signal (because each concatemer can bind many fluorescent probes). By designing different DNA sequences for different antibodies, multiple targets can be detected. The amplification increases sensitivity, enabling detection of low-abundance proteins which might be missed by direct IF. The DNA “imager” strands can also carry unique labels (like a sequence with a reporter for mass cytometry, see SABER-IMC below).

Immuno-SABER has been applied to immunofluorescence imaging where sensitivity is critical – e.g., detecting weakly expressed markers in tissues. It’s been used in multiplexed cell imaging, often in combination with cyclic methods: one can stain all antibodies (with DNA tags) at once, then sequentially introduce different fluorescent imagers to read out different targets. In a recent notable application, immuno-SABER was combined with Imaging Mass Cytometry (SABER-IMC) to amplify and detect 18 low-abundance immune markers in tumor tissues that standard IMC couldn’t see.

DEI (DNA Exchange Imaging)

DNA Exchange Imaging (DEI) is a high-speed multiplexed imaging method that, like CODEX, utilizes DNA-barcoded antibodies for spatial protein profiling. However, instead of using a microfluidics system to flow in and remove fluorescent DNA "reporter" strands (as in CODEX), DEI employs a rapid, manual exchange of fluorescent "imager" DNA strands. This approach eliminates the need for fluidics-based reagent delivery, significantly reducing cycle times.

The process begins with a single round of antibody staining, where all protein targets are labeled with DNA-barcoded antibodies. To visualize specific markers, a set of complementary fluorescently labeled DNA imager strands is manually introduced into the sample. These imager strands temporarily hybridize to their matching DNA barcodes on the antibodies, allowing fluorescence-based detection of a subset of targets.

Once a round of imaging is complete, a displacement buffer is quickly applied to remove the imager strands, erasing the fluorescence signal. A new set of fluorescent imager strands is then introduced, binding to a different subset of targets. Because the hybridization and displacement of short DNA oligos occur rapidly—on the order of seconds to minutes—each imaging cycle can be completed in under 10 minutes. This makes DEI much faster than CODEX, which relies on slower fluidics-controlled hybridization and wash steps.

Unlike some multiplexing methods that require antibody stripping and re-staining, DEI retains the same set of antibodies throughout the entire experiment. The only molecules being exchanged are the fluorescent imager strands, making it a non-destructive, high-throughput imaging technique. DEI has been demonstrated on both diffraction-limited and super-resolution microscopes (STED, SIM, PAINT), making it highly adaptable for various imaging applications.

However, DEI is currently mostly in the academic domain as a published method. It doesn’t have a turnkey commercial system like CODEX, so adoption requires setting up the workflow oneself.

DEI was applied to cultured cells and tissue sections to image a variety of cell structures. Wang et al. (2017) used it to map proteins in neurons and breast cancer cells, showing it could resolve ~10 synaptic proteins in cultured neurons and even do protein–protein colocalization analysis in situ. It has been used in super-resolution mode to study nanoclusters of proteins by exchanging DNA-PAINT probes, enabling multiplexed super-resolved imaging of cell structures at 20nm resolution.

HCR (Hybridization Chain Reaction) - First adapted for protein detection around 2017

Hybridization Chain Reaction (HCR) represents an elegant DNA-based signal amplification strategy that has been adapted for spatial protein detection. Originally developed for nucleic acid visualization, HCR employs a triggered chain reaction of DNA hybridization events to create long, fluorophore-labeled DNA polymers at sites where proteins of interest are located. The method begins with antibodies conjugated to DNA initiator sequences that can trigger the HCR amplification process.

When these DNA-tagged antibodies bind to their target proteins in tissue, they present initiator sequences that serve as nucleation sites for HCR amplification. The amplification involves two metastable DNA hairpin molecules that, upon encountering the initiator sequence, undergo a cascade of hybridization events, creating a long polymer of interlocked DNA strands. Each of these strands can carry multiple fluorophores, dramatically amplifying the signal from a single antibody-binding event. This amplification strategy enables detection of low-abundance proteins that might be missed by conventional immunofluorescence methods.

The programmable nature of DNA interactions allows for multiple orthogonal HCR amplification systems to operate simultaneously, enabling multiplexed protein detection with minimal crosstalk. In the seminal adaptation for protein detection by Koos et al., researchers demonstrated 5-10 fold signal enhancement compared to traditional methods, revealing previously undetectable protein expression patterns in complex tissues.

in situ Hybridization Chain Reaction (isHCR) (2020)

In situ Hybridization Chain Reaction (isHCR) is a specialized adaptation of HCR designed for in situ detection of biomolecules within intact tissues and cells. It follows the same core principle as HCR but is optimized for spatially resolved imaging of mRNA transcripts or protein targets in complex biological environments.

In isHCR, DNA or RNA probes are hybridized to target molecules directly within fixed tissues or cells. After washing away unbound probes, the HCR amplification process begins, generating a localized polymer chain at the site of the target. Since isHCR does not require enzymatic reactions or extensive sample processing, it is well-suited for tissue sections, whole-mount imaging, and even live-cell applications.

2. Sequencing-based Methods

Sequencing-based spatial proteomics represents a different approach that uses next-generation sequencing (NGS) rather than fluorescence microscopy to map protein distributions within tissues. These methods translate protein identity into DNA sequences through barcoded antibodies that bind specific targets in tissue sections. By linking these protein-identifying barcodes to spatial coordinate information, the technology creates molecular records that can be sequenced and computationally reconstructed into comprehensive protein maps. This NGS-based readout strategy overcomes the well-known spectral limitations of fluorescence methods, enabling simultaneous detection of hundreds to thousands of proteins.

Spatial CITE-seq

Spatial Cellular Indexing of Transcriptomes and Epitopes by Sequencing (Spatial CITE-seq) is an advanced multi-omics method that combines spatial transcriptomics with surface protein profiling in intact tissues. It builds upon CITE-seq (Cellular Indexing of Transcriptomes and Epitopes by sequencing), a technique that uses oligonucleotide-barcoded antibodies to simultaneously measure protein expression and mRNA levels in single cells.

First, a thinly sectioned tissue sample is placed onto a spatial transcriptomics slide or microarray, typically one that contains a spatially barcoded RNA capture surface (e.g., AtlasXOmics DBiT-seq’s spatial barcoding kit). The sample is then incubated with a panel of DNA-barcoded antibodies, each targeting a specific surface protein. These antibodies allow for protein-level detection in addition to RNA sequencing. The tissue is permeabilized to release mRNAs, which bind to the spatially barcoded oligonucleotides on the slide. Simultaneously, the DNA-barcoded antibody tags remain associated with their respective cell locations.The mRNA and antibody-derived DNA barcodes are sequenced together using high-throughput sequencing. The spatial barcodes enable mapping of transcriptomic and proteomic data back to their original tissue locations. Finally, bioinformatics pipelines integrate these datasets to create a high-dimensional spatial map of both gene and protein expression.

The Spatial CITE-seq implementation, described by Liu et al. in 2023, was able to measure 273 proteins and transcriptome in human tissues, revealing spatially distinct germinal center reactions in tonsil and early immune activation in skin at the Coronavirus Disease 2019 mRNA vaccine injection site.

Molecular Pixelation

Immune cells play a fundamental role in disease mechanisms, driving inflammation, immune surveillance, and responses to infections and cancer. Understanding their behavior at the molecular level is critical for developing targeted therapies.

As such, Molecular Pixelation (MPX) is designed to provide a deep spatial analysis of immune cell surface proteins and their spatial interrelationships.

The workflow is based on a simple kit-based chemistry that can be performed in solution, eliminating the need for specialized instruments or single-cell compartmentalization. The process begins with peripheral blood mononuclear cells (PBMCs) or other immune cell types, which are first fixed with paraformaldehyde (PFA) to lock their spatial organization in place.

Cells are then incubated with antibody-oligo conjugates (AOCs) that are optimized and validated to bind specific cell surface proteins. These conjugates serve as the foundation for spatial barcoding, allowing precise molecular mapping of protein locations on each cell.

The core of MPX is the Pixelation process, which is conducted in two iterative steps. In “Pixelation A”, DNA-based molecular pixels—nanometer-sized repeat concatemers containing unique molecular identifiers (UMIs)—are introduced into the reaction. These molecular pixels hybridize with the antibody-conjugated oligonucleotide sequences, forming spatially localized DNA zones (Zone A) where specific proteins reside.

The process is repeated in Pixelation B, forming Zone B, which captures additional layers of protein spatial distribution. This results in a three-dimensional molecular mesh encoding protein positions relative to one another within each individual cell.

At the end of the reaction, the DNA-labeled proteins are converted into a PCR-amplifiable product, which is then sequenced using a next-generation sequencing (NGS) platform. Each single cell generates approximately 120,000 sequencing reads, providing an unprecedented level of spatial proteomic detail. The data is analyzed using Pixelgen’s Pixelator software, which reconstructs protein distributions, identifies cellular identities, and extracts key spatial metrics.

Unlike traditional flow cytometry, MPX’s software enables three key types of spatial proteomic analysis:

1. Protein abundance: Similar to flow cytometry or CyTOF, measuring the overall expression levels of proteins per cell.

2. Individual protein spatial distribution: Quantifying whether a single protein is randomly distributed, polarized, or clustered within the cell.

   • Polarity scores indicate whether a protein is organized (<0), clustered (>0), or randomly dispersed (=0).

   • Example: T cells stimulated via CD3 show polarized CD3 localization on one end of the cell.

3. Protein colocalization analysis: Determining how different proteins are positioned relative to each other on the same cell.

   • Colocalization scores range from <0 (segregated proteins) to >0 (proteins physically co-localized).

   • Example: Rituximab treatment of RAJI cancer cells polarizes CD20, a mechanism critical for CAR-T therapy and antibody-based drugs.

MPX provides spatial proteomic insights for up to 1,000 single cells per reaction, detecting 80 proteins per cell, with 80×80 colocalization events mapped per dataset. This enables researchers to study dynamic changes in the surface proteome in response to biological stimuli. A key example is T cell migration, where MPX reveals the polarized relocalization of CD50 when T cells move along an ICAM-1 coated surface with chemokine stimulation.

One of the most intriguing discoveries using MPX came from a study on CD4 T cells, where researchers compared stimulated vs. unstimulated cells. This revealed previously uncharacterized changes in the spatial organization of CD proteins, opening new avenues for immune system research and therapeutic targeting.

DSP

A well-known option for target region of interest (ROI)-based sequencing is GeoMx, which can profile the whole transcriptome and 600+ proteins simultaneously.

In DSP, tissue sections are stained with a cocktail of either oligonucleotide-barcoded antibodies or RNA hybridization probes, each probe carrying a unique DNA barcode and a photocleavable linker. The tissue is placed in the instrument, and regions of interest (ROIs) are selected (either by manual selection, predefined grid, or automated segmentation). UV light is then directed to each ROI in turn, which cleaves the DNA barcodes from the probes only in that illuminated area. These released DNA barcodes are collected and counted (either by next-generation sequencing or NanoString optical barcoding instrument nCounter) to quantify the expression of targets in that ROI.

In cancer research, an interesting discovery came from applying DSP to melanoma samples, where researchers identified distinct immune contextures that predicted response to immunotherapy. The original work by Merritt et al. revealed that effective anti-PD-1 therapy was associated not simply with T cell abundance, but with specific spatial arrangements of PD-1+ T cells relative to PD-L1-expressing tumor cells and myeloid populations, highlighting the importance of spatial proximity in immune checkpoint biology. Further, DSP analysis of COVID-19 lung tissues (Desai et. al, 2020) provided crucial insights into the pathogenesis of severe disease, revealing specialized damage-associated molecular niches where virus-infected cells exhibited distinct protein signatures that orchestrated aberrant immune responses. These spatially-restricted interaction zones helped explain the paradoxical coexistence of immune hyperactivation and immunosuppression observed in severe cases.

3. Mass Spectrometry-based Methods

Mass spectrometry (MS) provides a powerful means to analyze proteins within intact tissues, adding a complementary perspective to imaging- and sequencing-based spatial methods. Unlike conventional immunofluorescence imaging, which relies on a limited number of antibody labels, MS can detect hundreds of molecules in tissue without any labels or a priori target selection. This label-free capability allows MS-based spatial proteomics to map a broad molecular landscape (proteins, peptides, metabolites) in a single experiment, avoiding the spectral overlap and throughput limitations of fluorescence microscopy.

Mass spectrometry-based spatial proteomics approaches fall into two broad categories:

• Label-free methods: These analyze tissues for their intrinsic protein content by ionizing molecules and measuring their mass-to-charge signatures

• Metal-tagged methods: These combine immunohistochemistry with MS detection, using antibodies carrying metal isotopes instead of fluorophores

Label-free approaches are essentially untargeted (discovery-driven, detecting whatever proteins are present above the detection limit), whereas metal-tagged assays are targeted (each metal reporter corresponds to a specific antibody-labeled protein).

3.a. Label-Free Mass Spectrometry Approaches

Label-free spatial proteomics uses the inherent mass and charge properties of proteins or peptides to identify and map them in tissue. In these approaches, proteins from a tissue section are directly ionized and detected by the mass spectrometer, producing spectra of mass-to-charge features that correspond to different biomolecules. Because no labeling is needed, label-free MS can simultaneously detect a vast array of molecules, providing an unbiased view of the tissue's molecular composition.

There are two main modes to perform label-free spatial proteomics: LC-MS/MS and Imaging Mass Cytometry.

LC-MS/MS

Small regions of tissue (such as microdissected cells) are excised or isolated, then processed for conventional proteomic analysis. The proteins are typically digested into peptides and separated by liquid chromatography, followed by tandem MS identification. This workflow yields a detailed proteomic profile of specific locations, often identifying hundreds or thousands of proteins from a tiny pieces of tissue. Its strength lies in sensitivity and depth, but it's relatively low-throughput spatially since each location must be sampled and run separately.

Many technologies have emerged to increase the throughput of LC-MS. Examples include MASP, scDVP, and LCM-SISPROT.

LCM-SISPROT

In 2018, Xu et al. introduced LCM-SISPROT, a spatial proteomics approach designed to analyze specific pathological regions based on H&E staining. The method begins with H&E-stained tissue slides, which are carefully examined by an expert pathologist to assess cellular and morphological features. Using these annotations, regions of interest (ROIs) are precisely microdissected with a Leica LMD system for downstream proteomic analysis.

A key innovation of LCM-SISPROT is its two-stage Spintip-based processing, which effectively removes H&E staining dyes while preserving protein integrity. This allows for the extraction of spatially distinct proteomes from different cell types within the same tissue section, including enterocytes, lymphocytes, and smooth muscle cells. By applying this method across four consecutive tissue sections, Xu et al. successfully generated spatially resolved proteomic profiles at a micrometer scale. Notably, they identified over 500 proteins from a minute colon cancer tissue sample measuring just 0.1 mm² in area and 10 µm thick.

MASP: Micro-Scaffold Assisted Spatial Proteomics

In 2022, Ma et al. developed Micro-Scaffold Assisted Spatial Proteomics (MASP), an approach that enhances spatial proteomics by integrating 3D-printed micro-scaffolds for localized protein capture, preservation, and enzymatic processing. Unlike conventional spatial proteomics techniques that rely on laser microdissection or direct tissue digestion, MASP physically compartmentalizes tissue regions within a structured scaffold, ensuring that proteins are processed in situ while maintaining their spatial organization. The result is a mouse brain tissue that was divided into hundreds of 400 µm × 400 µm voxels, with over 5000 proteins detected in each voxel. Using this approach, the authors developed a spatial proteomic atlas of Alzheimer’s disease in a mouse cerebral cortex.

scDVP: Single-Cell Deep Visual Proteomics

Single-Cell Deep Visual Proteomics (scDVP) (Rosenberger et. al, 2023) is an advanced approach that combines high-resolution imaging with single-cell proteomics, allowing researchers to identify and quantify proteins in individual cells while retaining information about each cell’s location and phenotype. This method, often referred to simply as Deep Visual Proteomics (DVP), integrates the strengths of microscopy and mass spectrometry into one pipeline.

First, a tissue section (or cell culture) is imaged in detail – for example, using microscopy with fluorescence or histological stains – to visualize cell morphology and context.

Artificial intelligence-driven image analysis is then used to segment the image and recognize cells or regions of interest based on their appearance (for instance, cell shape, size, or marker intensity)​.

scDVP next employs automated laser capture microdissection to physically isolate those target cells, one cell at a time, directly from the tissue slide. Each isolated single cell (or nucleus) is subjected to an ultrasensitive sample preparation and MS analysis, typically using optimized single-cell proteomics workflows (e.g. specialized lysis and nanoLC-MS).

Finally, the proteomic data from each cell is linked back to its imaging data and location. This yields a rich output: one can correlate microscopic phenotypes (such as cell type, histological features, or spatial position) with the deep proteome of the same single cell. As a result, scDVP was able to identify a maximum of 2769 proteins in a single cell.

Imaging Mass Spectrometry (IMS)

Unlike LC-MS/MS methods, which require microdissection and protein digestion before tandem MS analyses, IMS acquires data across an entire tissue section in a rasterized fashion, yielding a map of ion signals at each x,y coordinate.

The tissue is analyzed intact on a slide: the instrument's probe moves across the sample, analyzing countless tiny spots (pixels) and recording the intensity of each mass-to-charge ratio at each spot. The result is a set of ion images, where each image represents the spatial distribution of a particular molecule in the tissue. This approach preserves continuous spatial information similarly to a microscope image, albeit with the rich molecular content that MS provides.

Two Approaches for IMS: Top-down vs. Bottom-up

IMS can be further divided into two key approaches: top-down and bottoms-up.

Top-down IMS

Top-down imaging mass spectrometry (IMS) detects intact biomolecules—proteins, lipids, and metabolites—directly from tissue without prior digestion. This approach preserves molecular structures and post-translational modifications (PTMs), making it ideal for mapping spatial distributions of biomolecules in situ.

MALDI-MSI

A prime example is MALDI-MSI (Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry Imaging), which has become one of the most widely used label-free IMS methods. In MALDI-MSI, the tissue is coated with a matrix compound and a laser beam is fired in a grid pattern across the section. The laser's energy, absorbed by the matrix, causes nearby biomolecules to desorb and ionize. MALDI can detect a range of biomolecules (from small metabolites up to proteins of ~20–30 kDa) with high sensitivity. Recent advancements such as MALDI-2, which introduces a secondary laser pulse to enhance ionization, have further improved sensitivity and resolution.

DESI

Other top-down IMS techniques operate on similar principles but with different ionization mechanisms. For instance, DESI (Desorption Electrospray Ionization) uses a stream of charged micro-droplets directed at the tissue; as the solvent impacts the surface, it lifts off molecules, which then enter the MS.

A variant called nanoDESI uses a nanoscale solvent junction for finer control: two tiny capillaries create a micro-droplet that scans the surface, achieving improved sensitivity and lateral resolution over standard DESI​. Indeed, nanoDESI has demonstrated spatial resolution below 10 µm for imaging lipids in tissue, a significant improvement given that older DESI setups were often limited to ~100 µm scale​

LESA

LESA (Liquid Extraction Surface Analysis) is another powerful tool in top-down IMS. In this technique, a robotic pipette deposits a small solvent droplet onto a tissue section, then re-aspirates it to extract biomolecules for mass spectrometry analysis. LESA is particularly advantageous for extracting intact proteins and metabolites while allowing tandem MS (MS/MS) analysis for deeper molecular characterization. Although LESA lacks the high spatial resolution of MALDI or DESI, it enables solution-phase proteomics with enhanced analytical depth.

Overall, top-down IMS techniques provide diverse capabilities for molecular imaging. MALDI and nanoDESI enable high-resolution, in situ imaging, making them suitable for lipidomics and small-molecule analysis. In contrast, LESA offers higher analytical depth at the cost of spatial resolution, making it well-suited for proteomics applications that require detailed molecular characterization.

Bottom-up IMS

In contrast to top-down IMS, bottom-up imaging mass spectrometry (IMS) incorporates an enzymatic digestion step to analyze peptides rather than intact proteins in situ. In these workflows, a tissue section is treated with a protease—typically trypsin—to cleave proteins into peptides directly on the slide. These peptides, being smaller and more amenable to tandem mass spectrometry (MS/MS) sequencing, produce more identifiable fragment ions than intact proteins, improving protein identification and spatial mapping. While this method enhances the number of proteins detected, it does not retain information on intact protein structures or post-translational modifications as effectively as top-down approaches.

The same ionization technologies used in top-down IMS—MALDI, DESI, nanoDESI, and LESA—are also applicable in bottom-up IMS, but with a focus on detecting peptides rather than intact biomolecules. MALDI-MSI, for example, remains a core technique, but instead of detecting whole proteins, it ionizes digested peptides for spatially resolved MS/MS sequencing. This allows researchers to generate peptide maps of tissue sections, significantly improving protein identification over direct intact protein detection.

3.b. Metal-Tagged Mass Spectrometry Approaches

Metal-labeled imaging approaches use antibodies tagged with metal isotopes instead of fluorophores. The readout is performed by mass spectrometry, allowing dozens of different metal tags to be detected in parallel without spectral overlap. Two main implementations exist: laser ablation inductively coupled plasma mass spec (Imaging Mass Cytometry, CyTOF-based) and ion beam imaging mass spec (MIBI).

1. Imaging Mass Cytometry (IMC)

IMC, developed by Fluidigm (now Standard BioTools), adapts CyTOF mass cytometry for imaging applications. A focused laser (UV ablation) scans across the tissue, vaporizing microscopic regions (~1 µm spots). The resulting plume of ablated material is carried into the CyTOF time-of-flight (TOF) mass spectrometer, where metal tags are detected and quantified. IMC provides single-cell resolution, allowing researchers to visualize cellular organization and protein expression in complex tissues. The technique comes with the advantage of high multiplexing (~40 markers), single-cell resolution, and no spectral interference. On the other hand, the laser fully ablates the sample, preventing rescanning or additional downstream analysis.

2. Multiplexed Ion Beam Imaging (MIBI)

MIBI, developed by IonPath, employs an ion beam (e.g., oxygen duoplasmatron) instead of a laser to sputter the surface of the tissue pixel by pixel. This dislodges metal-labeled ions, which are then analyzed by a TOF mass spectrometer, similar to IMC. Unlike IMC, MIBI allows multiple scans of the same region, as only a thin layer of tissue is removed per pass. One advantage of MIBI is the flexible resolution tuning (from whole-tissue to subcellular scales), rescan capability, and reduced sample consumption. On the other hand, is also has slightly slower acquisition speed compared to IMC due to pixel-by-pixel sputtering.

Both techniques generate high-dimensional tissue maps, where each metal isotope corresponds to a protein, and its intensity at each XY coordinate represents local abundance. The resulting datasets resemble multiplexed immunofluorescence images but without the constraints of fluorescence-based methods.

Over the past decade, many advances have been built on top of these two foundational technologies. We proceed by reviewing these developments below:

SABER-IMC

SABER-IMC is a hybrid method that combines DNA-mediated signal amplification (immuno-SABER) with Imaging Mass Cytometry (IMC).

In standard IMC, one limitation is that very low-abundance proteins may fall below detection threshold. SABER-IMC addresses this by first performing immuno-SABER: antibodies for selected low-abundance targets carry DNA concatemers (via the SABER method, see above) to amplify the number of DNA binding sites. Next, instead of fluorescent imagers, those DNA concatemers are hybridized with isotope-labeled oligonucleotide probes. These metal-containing probes bind in large numbers to each concatemer, effectively loading lots of metal atoms at each target site. Then the sample is run on an IMC instrument: a laser ablates the tissue and the plume is analyzed by a CyTOF mass cytometer, detecting both the standard metal-tagged antibody signals and the DNA-bound isotope signals. This way, markers enhanced by SABER show greatly boosted signal counts in the mass cytometry readout. SABER-IMC typically is used in a mode where some markers (maybe 10–20) are amplified and others (20–30) are done conventionally, achieving ~30–50-plex in total.

SABER-IMC has been applied to human tumor samples to reveal biomarkers that were previously undetectable by IMC alone. For example, in a melanoma study, 18 low-expression proteins (such as certain T cell co-receptors and ligand proteins) were successfully imaged with SABER amplification, alongside 20 regular markers. This allowed identification of immune cell subpopulations that would have been missed by standard IMC.

HD-MIBI

HD-MIBI is an advancement of MIBI that achieves much higher spatial resolution (down to ~30–50 nm) at the cost of multiplexing or area size. This is achieved by using a finer primary ion beam and optimized instrument settings to get very small sputter spots. As a result, HD-MIBI can zoom in on tiny structures (for example, subcellular distributions of drugs or signaling molecules) that standard MIBI cannot resolve. The time-of-flight mass spectrometer still detects multiple isotopes, but because of the much smaller beam and slower sputter, typically fewer markers (maybe 6–8) are practical and areas imaged are relatively small.

HD-MIBI has been used to study subcellular localization of molecules like pharmaceuticals within cells in tissue. For instance, it can visualize a drug and a couple of proteins in a cell to see if the drug is penetrating the nucleus or stuck in vesicles. It’s also useful for examining microdomains in tissues, e.g., membrane protein clusters, synapses, etc., where 50 nm resolution illuminates spatial relationships beyond light microscopy.

Metal-tagged spatial proteomics marries the specificity of antibody-based targeting with the multiplexing and quantitative abilities of mass spectrometry. In these methods, the spatial localization comes from antibodies binding their target proteins in the tissue, and the readout comes from detecting attached metal isotopes via MS. The general workflow is analogous to an immunohistochemistry (IHC) stain, but instead of chromogenic or fluorescent labels, each antibody is conjugated to a unique metal isotope tag (often a lanthanide rare-earth metal). After the tissue is incubated with a cocktail of, say, 30 or 40 different metal-labeled antibodies, it is introduced into an imaging mass spectrometer that can detect those metals with spatial fidelity​

Comparison between Commercial Solutions

Below, we present an overview of 14 spatial proteomics products on the market and a comparison of their attributes.

Number of Targets & Throughput

All products can be loosely grouped into three categories:

One key consideration when choosing low vs. high-plex spatial proteomics methods is the type of study and biological questions one wishes to carry out.

Low-plex platforms are ideal for targeted studies, such as validating specific biomarkers or conducting routine clinical diagnostics at fast speed. For example, Akoya Biosciences’ PhenoImager can process a whole-slide 6-plex in <12 minutes and hundreds of samples per week with high reproducibility and robust data quality. This is especially crucial for high-throughput translational studies and clinical trials where biomarker signatures need to be analyzed across large cohorts of patient samples. InSituPlex, when used with Leica Biosystems' BOND RX automated research stainer, allows for processing up to 30 tissue sections in <6 hours.

Medium-plex platforms, which support 21 to 60 markers, provide a balance between depth and throughput, making them suitable for tumor microenvironment studies, immune-oncology research, and functional tissue analysis. These systems allow for a more comprehensive view of cell interactions while keeping data complexity manageable.

For studies requiring deep profiling of cellular interactions and molecular states, high-plex platforms provide the most comprehensive insights. These systems support over 100 markers, allowing researchers to explore vast biological landscapes in a single experiment. PhenoCycler®-Fusion 2.0 and MALDI HiPLEX-IHC can both detect over 100 biomarkers, while the Pixelgen Proxiome Kit, Immuno 155 is specifically designed for cell surface marker colocalization, featuring 155 pre-designed surface protein markers. MACSima expands on this, enabling the simultaneous detection of over 200 protein markers and 27 RNA targets, offering multiomic capabilities within a single assay. At the highest end of the spectrum, Microscoop® Mint and MALDI GUIDED SPATIALOMX® extend detection to 1,000+ markers, making them particularly valuable for hypothesis-free research. On the other hand, these systems may take more than a single day, sometimes multiple days, to go from raw samples to data.

Panel Flexibility and Customization for Antibodies

A major challenge in antibody-based spatial proteomics is ensuring specificity and reproducibility across multiple markers. Poorly validated antibodies can cause cross-reactivity and signal interference, making pre-defined, validated panels crucial for reliable results, especially in immuno-oncology (IO) and tissue-specific studies.

Many platforms offer ready-to-use antibody panels optimized for FFPE and fresh-frozen tissues. Aurora 3D Spatial Biology Solution is has a predefined panel specifically for dermatology, for example, while Pixelgen Proxiome Kit, Immuno 155 specializes in immune cell surface marker colocalization with 155 pre-designed markers.

We provided a summary table above to link and redirect eaders to specific companies’ website. It is important to check out if your tissue, species, and biomarkers of interest belong to the list of optimized, pre-defined panels at each company. We recommend evaluating these panels early in the experimental design process to streamline workflow setup and minimize validation challenges.

Tissue Types

The majority of providers support fresh/frozen, FFPE, and tissue microarrays. Pixelgen requires sample inputs to be fixed cells in suspension originating from PBMCs, BM, dissociated organoids, or fresh frozen tissue.

The Aurora 3D™ Spatial Biology Solution by Alpenglow Biosciences is currently the only platform that enables comprehensive 3D protein profiling of intact tissues, while most other platforms work with tissue slices that are 5-10 microns thick.

Maximum Scan Area per Run

(Caveat: We couldn't find scan area and slide capacity details for Aurora 3D Spatial Biology Solution, Microscoop® Mint, MALDI HiPLEX-IHC, MALDI GUIDED SPATIALOMX®, and InSituPlex®, so our comparison includes only platforms with available data.)

The PhenoImager® HT 2.0 by Akoya Biosciences leads in maximum scan area per run, reaching 132,240 mm². This is due to its large scan area per slide (ranging from 15 mm x 15 mm to 29 mm x 57 mm) and its high-throughput capacity, supporting up to 80 slides per run with continuous loading technology for low-plex panels.

Following this, the Hyperion XTi™ System reaches 37,120 mm², balancing medium-plex analysis (42 markers) with high-throughput capabilities when paired with its optional slide loader. Cell DIVE, from Leica Microsystems, comes in third with 13,500 mm², particularly when integrated with the BioAssemblyBot 200 from Advanced Solutions Life Sciences, which enhances its automation and scanning efficiency.

Other platforms, such as Orion™, MACSima, and PhenoCycler®-Fusion 2.0, offer smaller scan areas but provide higher plex options, indicating a trade-off between scan coverage and marker depth.

Cost

Initial capital equipment investment:

• High-end systems ($500,000-$1,000,000+): Mass spectrometry-based platforms like Imaging Mass Cytometry (IMC) and Multiplexed Ion Beam Imaging (MIBI) represent the highest initial investments, requiring specialized instrumentation and dedicated facility space. These systems are typically found in well-funded research institutions and pharmaceutical companies.

• Mid-range systems ($250,000-$500,000): High-plex fluorescence imaging technologies such as advanced cyclic immunofluorescence platforms fall into this category, offering a balance between multiplexing capability and cost.

• Entry-level systems ($100,000-$250,000): Lower-plex imaging systems and attachments to existing microscopy platforms provide more accessible entry points for spatial proteomics.

Consumables and operational cost:

The per-sample expenses follow a consistent pattern across platforms:

• Low-plex approaches (5-10 markers): $50-200 per sample

• Medium-plex approaches (20-60 markers): $200-800 per sample

• High-plex approaches (100+ markers): $800-2,000+ per sample

Service-based alternatives:

Many institutions provide spatial proteomics as a service model, offering a more cost-effective approach for groups with limited access to capital funding or specialized expertise. Core facilities typically charge: (1) base fee per sample type/preparation, (2) per-marker fees that increase with multiplexing level, and (3) addditional fees for custom antibody conjugation, validation, and advanced analysis

The potential hidden costs of antibody validation

Antibody validation often consumes unexpected time and budget in spatial proteomics projects. Panel development becomes progressively complex with each additional marker due to challenges in antibody compatibility, cross-reactivity, and epitope stability. The process involves costly steps including:

• Sourcing appropriate antibodies (1-4 weeks)

• Individual validation in specific tissue contexts (3-6 weeks per tissue type)

• Technology-specific optimization, as antibodies validated for traditional IHC often perform differently on spatial platforms

• For cyclic methods, confirming stability across multiple staining-imaging-stripping cycles

• Iterative panel adjustments when changing one antibody affects others' performance

Pre-validated commercial panels can significantly reduce timelines but may be inadequate for novel tissue types, non-human organisms, or emerging disease biomarkers, often necessitating custom panel development that incur additional costs.

Multiomic Capabilities

Many platforms enable RNA detection through manual sequential workflows. The PhenoCycler®-Fusion 2.0 and PhenoImager® HT 2.0 preserve sample integrity, allowing subsequent H&E, IHC, or in situ RNA analysis. Bruker's CellScape employs EpicIF technology with white light to gently remove signals while maintaining epitope and tissue integrity, enabling downstream IF, RNA-ISH, and proximity ligation assays. Similarly, Cell DIVE's dye inactivation process leaves samples undamaged, permitting additional downstream applications.

Other platforms allow for automated RNA and protein co-detection on the same slide, which can decrease workflow length. For example, COMET, Hyperion XTi™ System, and InSituPlex all allow for co-detection of up to 12 RNAs when combined with ACD Bio-techne RNAScope® assays. MACSima is able to provide up to 27 RNAs during protein co-detection via a multi-layered workflow combining its proprietary MICS (protein detection) and RNAsky™ (RNA detection) technologies. First, for RNA detection, it uses padlock probes that hybridize directly to mRNA targets without cDNA synthesis, and amplifies signals via rolling circle amplification (RCA), generating discrete fluorescent puncta. After detection, the RNA probes are washed away and cyclic protein staining with fluorescent antibodies begins, enabling detection of up to 200+ protein markers.

While some platforms emphasize RNA detection, others focus on broader molecular profiling. MALDI HiPLEX-IHC and MALDI GUIDED SPATIALOMX®, for example, prioritize the detection of metabolites, lipids, proteins, glycans, and xenobiotics rather than RNA.

Future Perspectives

Spatial proteomics has transformed our understanding of protein biology by preserving the critical spatial context within intact tissues, revealing intricate protein-protein interactions and cellular neighborhoods impossible to detect with traditional methods. This review explores the three main technological branches that have evolved to address this challenge: fluorescence imaging-based methods that overcome spectral limitations through clever multiplexing strategies; sequencing-based approaches that translate protein identity into DNA barcodes for high-throughput readout; and mass spectrometry techniques that provide either label-free discovery or metal-tagged targeted analysis. The commercial landscape now offers solutions ranging from low-plex, high-throughput systems ideal for clinical translation to ultra-high-plex discovery platforms detecting over 100 protein markers simultaneously. What does the future of the field hold?

Increasing Number of Targets

We’re entering the era of ultra-high plexity. What used to be cutting-edge, 30 or 40 markers, is now just the baseline. Single-cell deep visual proteomics (scDVP), pioneered by Rosenberger et al., shattered expectations by identifying 2,769 proteins from a single cell. On the commercial side, Bruker announced a 1,000-plex upgrade to its GeoMx platform, marking one of the highest levels of tissue proteomics to date. Similarly, Element Biosciences introduced new capabilities to the AVITI24 system, offering users flexible protein customization with up to 88 proteins per assay using pre-designed or user-defined probes. These developments signal a shift toward deep profiling at scale, enabling both discovery-driven and targeted studies.

From Imaging to Sequencing

The spatial proteomics field is expanding beyond fluorescence and mass spec–based imaging to include sequencing-based readouts. These methods leverage barcoded antibodies to convert protein identity into DNA, enabling NGS-based quantification at high multiplexity without spectral constraints. Notably, STOmics’ Stereo-CITE Proteo-Transcriptomics Solution is currently the only commercial platform offering unbiased co-detection of whole transcriptome and hundreds of proteins on the same tissue section, at single-cell resolution. Another notable example is Molecular Pixelation, developed by Pixelgen. MPX uses DNA-barcoded antibodies and a novel “pixelation” chemistry to capture both protein abundance and spatial arrangement on the surface of single cells — all without imaging. In MPX, cells are stained with oligo-conjugated antibodies, then treated with synthetic DNA "pixels" that bind locally, creating nanoscale spatial maps of protein distributions. These are then sequenced, generating high-resolution data on protein localization, polarity, and co-localization across thousands of single cells per run.

Reference Atlases

The field isn’t just generating data — It’s building reference blueprints for human biology. Bruker Spatial Biology, in partnership with Weill Cornell Medicine, is expanding the Spatial Atlas of Human Anatomy (SAHA) initiative, a major effort to map 30 non-diseased organs across multicellular, single-cell, and subcellular levels in a diverse adult population. These high-resolution atlases will be made publicly available, supporting cross-institutional benchmarking and biological discovery.

Software

As spatial data grows in both dimensionality and scale, software becomes the linchpin for extracting biological meaning. Platforms differ widely in their computational maturity: commercial systems often provide integrated pipelines for cell segmentation, phenotyping, and neighborhood analysis, while academic tools (e.g., Squidpy, Giotto, Napari plugins) offer flexibility but require coding expertise. Critical software differentiators include:

• Cell segmentation accuracy (e.g., AI-assisted segmentation vs. classical watershed)

• Spatial statistics support and visualization (e.g., colocalization, spatial correlation of differential expressed genes)

• Multiomic integration tools (RNA + protein)

• Scalability

As public spatial reference atlases become more available, we expect software interoperability and standard formats (like OME-TIFF, AnnData) to be key enablers of cross-study integration and meta-analysis. Ultimately, the platforms that combine flexible assay design with scalable, interpretable software will become the backbone of translational spatial biology.

We're excited to see how the field evolves. Maybe one day we will be able to see the entire proteome of every single cell, at every single location of the human body.

This article is the second installment in a series investigating different “branches” of spatial technology. In part one, we took a look at sequencing-based technologies. We will focus on imaging-based methods.

This article will be organized by core concepts, methods by history and methods by attributes.

Please see this table of content to see the overall structure of the article and quickly jump to the part you’re interested in:

1. The technology tree

2. Core concepts

3. Methods by History

   1. Targeted In Situ Sequencing (ISS)

      1. First ISS method introduced in 2013

      2. ExSeq (2021): Reducing molecular crowding with expandable hydrogel in expansion microscopy

      3. STARmap (2018): Increasing signal-to-noise ratio and detection efficiency by removing the need for reverse transcription

      4. STARmap PLUS (2023): Enabling protein co-detection

      5. RIBOmap (2023): Detecting actively translating mRNAs

   2. Untargeted in situ sequencing

   3. Fluorescence In Situ Hybridization (FISH) methods

      1. Beyond smFISH: How seqFISH and MERFISH Solved These Problems

      2. How seqFISH Works: Expanding smFISH with Temporal Barcoding (2014)

      3. seqFISH+ (2019)

      4. MERFISH (2015): Multiplexed Error-Robust Fluorescence In Situ Hybridization

4. Methods by Properties

   1. Gene Panels: Customizability & Scale

   2. Throughput – Imaging Area Per Week

   3. Tissue Types Profiled: Beyond FFPE & Fresh Frozen

   4. Multiomic Capabilities: RNA & Protein Profiling

   5. Sensitivity

   6. Specificity: Accuracy of Gene Detection

   7. Species

5. Future developments

   1. Higher plexity towards whole transcriptome

   2. Live-cell imaging

   3. 3D atlas construction

   4. Unified ecosystem of verified computational tools

A Review of the Technology Tree

High-throughput spatial profiling technologies can be divided into two umbrella categories: sequencing-based and imaging-based methods. To help navigate these groups and contextualize their various branches, we’ve constructed a “technology tree”. For further exploration, we refer the reader to this online database to browse and filter technologies by different properties.

Sequencing-based methods involve strategies that ultimately use DNA sequencing to decode spatial information:

• Solid-phase capture: capturing molecules on barcoded surfaces (eg. ST/Visium)

• Deterministic barcoding: applying spatial barcodes directly to the tissue via microfluidics device (eg. DBiT-seq)

• Photo-activated barcoding: using light to activate spatial barcoding in specific tissue regions (eg. Light-seq)

Imaging-based methods fall into two main categories:

• In situ sequencing (ISS) involves hybridizing probes to target transcripts, amplifying the signal through techniques like rolling circle amplification, and then sequentially reading nucleotide sequences using iterative rounds of fluorescent labeling and imaging.

• In situ hybridization (ISH), on the other hand, detects specific RNA or DNA sequences by hybridizing many fluorescently labeled probes directly to target molecules, enabling spatial localization.

This article focuses only on imaging-based, transcriptomic technologies and even then gives a cursory overview of this material. See our previous post on sequencing-based methods.

Imaging-based Methods

Core Concepts

It is useful to examine imaging technologies in comparison to Sequencing-based methods use NGS to decode spatial barcodes. Imaging-based technologies use high-resolution microscopy to directly visualize and quantify RNAs and proteins in their native tissue context.

Sequencing-based approaches excel at unbiased, whole-transcriptome profiling but can struggle with detection efficiency and sensitivity due to inefficiencies in cDNA conversion, amplification bias, and molecular dropout. In contrast, imaging-based technologies deliver higher detection efficiency and single-molecule resolution, making them superior for capturing rare transcripts and preserving spatial organization with minimal molecular loss. However, they are constrained by multiplexing limits—currently, the highest-plex imaging methods can detect up to 1,000 genes, a fraction of the ~18,000 expressed in the human genome.

To navigate the landscape, we divide imaging-based methods into three main categories:

• Targeted in situ sequencing

• Untargeted in situ sequencing

• Fluorescence in situ hybridization (FISH)

Methods by History

Targeted In Situ Sequencing (ISS)

First ISS method introduced in 2013

The original paper for in situ sequencing in preserved cells and tissues was published in 2013 by Ke et al. The process begins by converting messenger RNA (mRNA) into complementary DNA (cDNA) using reverse transcription, followed by the degradation of the original mRNA.

A specialized padlock probe, designed to match the cDNA, hybridizes to it but leaves a small gap, which is then filled and sealed by DNA polymerase and ligase, forming a circular DNA structure. This DNA circle undergoes rolling circle amplification (RCA), generating multiple copies that create a bright fluorescent signal detectable under a microscope.

To read the genetic sequence, fluorescently labeled probes—each specific to a DNA base (A, C, G, or T)—are introduced, binding to their complementary bases. The sample is then imaged to capture the fluorescence pattern, and this process is repeated iteratively to determine the entire sequence. Computational analysis deciphers the fluorescence signals, reconstructing a spatial gene expression map that reveals how different genes are expressed within tissues.

One common disadvantage of ISS is low gene detection efficiency (likely in the order of 5%), which can be possibly attributed to low efficiency of in situ cDNA synthesis. Reverse transcription may fail to capture all mRNA molecules, leading to incomplete cDNA generation. Structural barriers in tissue, such as cross-linking in fixed samples (e.g., FFPE tissues), can hinder enzyme access to RNA.

ExSeq (2021): Reducing molecular crowding with expandable hydrogel in expansion microscopy

Targeted expansion sequencing (ExSeq) (2021 by Alon et al. at Boyden’s lab) addresses the challenges of low gene detection efficiency in traditional in situ sequencing (ISS) by integrating expansion microscopy (ExM) with ISS techniques.

The technique works by first introducing specific chemical anchors are introduced to bind RNA molecules within the tissue. This step ensures that RNA remains attached to the tissue matrix during subsequent processing. The tissue is embedded in a swellable polymer gel, typically composed of acrylamide and other monomers, permeating the tissue and linking to the anchored RNA molecules. The embedded tissue is exposed to water, causing the hydrogel to swell isotropically (equally in all directions).

This expansion physically separates the RNA molecules, enhancing their spatial resolution and accessibility for subsequent reverse transcription into complementary DNA (cDNA). The process continues with rolling circle amplification (RCA) and in situ sequencing to generate optical barcodes corresponding to the RNA identity.

Targeted ExSeq has been used to detect tens to hundreds of genes in the mouse primary visual cortex, hippocampus and human metastatic breast cancer biopsy samples, with ~40% detection efficiency.

STARmap (2018): Increasing signal-to-noise ratio and detection efficiency by removing the need for reverse transcription

STARmap is another improvement from the original in situ sequencing technique to both increase the signal-to-noise ratio and gene detection efficiency.

STARmap achieves a detection efficiency of 73%, significantly higher than earlier in situ sequencing methods (~5–10%), while profiling 160-1000 genes over six imaging cycles simultaneously. STARMAP also measured 28 genes across millimeter-scale volumes encompassing ~30,000 cells, revealing 3D patterning principles that jointly characterize a broad and diverse spectrum of cell types.

STARmap works based on four core innovations:

1. Targeted signal amplification with special SNAIL probe design.

2. Hydrogel chemistry to embed 3D tissue structure

3. Novel sequencing-by-ligation method called SEDAL

4. 3D imaging

SNAIL probe hybridization - No reverse transcription: Traditional in situ sequencing relies on inefficient reverse transcription of RNA to cDNA, limiting detection sensitivity. SNAIL eliminates this step by using a primer and a padlock probe that bind adjacent RNA regions. When both attach, the padlock probe circularizes and undergoes rolling-circle amplification, forming a DNA nanoball with multiple target sequence copies.

Embedding cDNAs in Hydrogel: The synthesized cDNA is embedded in a hydrogel matrix via in situ polymerization, preserving tissue architecture and immobilizing cDNA. A clearing step removes unbound molecules, enhancing optical transparency for downstream analysis.

In Situ Sequencing with SEDAL: Using SEDAL (STARmap’s sequencing-by-ligation method), a reading probe binds the single-stranded DNA template. A two-base decoding system assigns fluorescent labels to specific dinucleotide sequences (e.g., AA = blue, GC = red).

After ligation, fluorescence imaging captures base identity while unligated probes are washed away. Formamide treatment strips probes while preserving DNA amplicons, allowing sequencing cycles to repeat until the entire template is read.

STARmap PLUS (2023): Enabling protein co-detection

In 2023, STARmap PLUS was published as a method that combines high-resolution spatial transcriptomics with protein detection in the same tissue section. Proteins are labeled with primary antibodies, and cDNA amplicons, antibodies, and endogenous proteins are co-polymerized in a hydrogel. Fluorescent secondary antibodies and small-molecule dyes (e.g., X-34 for Aβ plaques) visualize protein localization.

Another notable improvement is the gene-coding barcode in DNA probes was expanded from 5 to 10 nucleotides, increasing coding capacity from 1,024 to over 20,000 genes, covering the whole transcriptome.

RIBOmap (2023): Detecting actively translating mRNAs

In the same year, RIBOmap was developed, allowing for the detection of actively translating mRNAs in intact cells and tissues. Unlike STARmap, which detects all mRNA molecules, RIBOmap employs a selective tri-probe design that specifically targets and amplifies ribosome-bound mRNAs, offering insights into protein synthesis rather than just mRNA presence.

This technique can profile thousands of genes simultaneously, with studies demonstrating the ability to profile 5,413 genes in mouse brain tissue. The development of Deep-RIBOmap further extends the technique's capabilities, allowing for 3D in situ quantification of translation activities within 200-μm thick tissue blocks.

These advancements enable RIBOmap to provide more comprehensive and insightful molecular maps of protein synthesis in the context of subcellular architecture, cell types, and tissue anatomy, offering a powerful tool for understanding translational regulation in various biological contexts.

Untargeted in situ sequencing

Fluorescent In Situ Sequencing (FISSEQ) is an untargeted spatial transcriptomics method that sequences RNA while preserving its spatial context in intact cells and tissues. Unlike targeted methods that rely on pre-designed panels of gene-specific probes, FISSEQ enables transcriptome-wide analysis without predefined targets.

This is achieved by using random hexamer primers to convert all RNA molecules into complementary DNA (cDNA).

The process begins with reverse transcription (RT), where fixed cells or tissues are permeabilized, and random hexamer primers convert RNA into cDNA. The original RNA is degraded to prevent interference, and the cDNA is circularized using CircLigase. To maintain spatial localization, cDNA is cross-linked to the cellular matrix using aminoallyl-dUTP and a chemical cross-linker. RCA then generates long, repeated DNA sequences, forming a stable 3D nucleic acid matrix that can be directly sequenced.

FISSEQ uses sequencing-by-ligation (SOLiD chemistry) to read out the sequences. Fluorescently labeled dinucleotide probes are iteratively ligated to the amplicons, and a confocal microscope captures images of the fluorescence patterns. Computational analysis, including image deconvolution and base calling, reconstructs gene expression patterns across tissue sections.

The technique is highly multiplexed but has a very low gene detection efficiency (~0.01%), mainly due to inefficiencies in cDNA synthesis and RCA.

Untargeted ExSeq combines modified FISSEQ protocols with expansion microscopy to improve efficiency, resulting in improved but likely still modest detection efficiency.

Fluorescence In Situ Hybridization (FISH) methods

Single-molecule fluorescence in situ hybridization (smFISH) is a technique used to visualize and quantify individual RNA molecules within fixed cells or tissue. It relies on fluorescently labeled DNA probes that hybridize to target RNA sequences, enabling high-resolution spatial gene expression mapping. Multiple probes bind a single transcript to amplify the signal while maintaining specificity. This is unlike in situ sequencing, which employs rolling circle amplification for signal enhancement.

However, despite its precision, smFISH has historically faced two major challenges: limited multiplexing and optical crowding.

1. Multiplexing constraints: Traditional fluorescence microscopes can only detect a few fluorophores at once, limiting researchers to visualizing only a handful of RNA species in a single experiment. What if scientists wanted to track thousands of genes simultaneously?

2. Optical crowding: In densely packed cellular environments, RNA molecules can cluster so tightly that fluorescence signals overlap, making it difficult to distinguish and count individual molecules.

Beyond smFISH: How seqFISH and MERFISH Solved These Problems

To overcome multiplexing and optical crowding, researchers developed seqFISH and MERFISH, which use barcoding strategies to massively increase RNA detection without spectral overlap:

seqFISH (Sequential FISH):

• Uses a temporal barcoding approach, where the same fluorophores are reused in multiple hybridization rounds. This allows for the detection of hundreds to thousands of genes in a single experiment while avoiding fluorescence spectral overlap.

MERFISH (Multiplexed Error-Robust FISH):

• Uses combinatorial barcoding with built-in error correction, enabling the detection of over 10,000 different RNA species while preventing misidentification.

We will explore how each method works below.

How seqFISH Works: Expanding smFISH with Temporal Barcoding (2014)

smFISH detects RNA molecules by hybridizing fluorescently labeled probes to their target sequences. Each gene typically requires a unique fluorophore, which limits the number of RNAs that can be simultaneously visualized to the number of spectrally distinct dyes available (~3-5).

On the other hand, instead of using a distinct fluorophore for each gene, seqFISH reuses a small set of fluorophores across multiple imaging cycles, allowing for the detection of thousands of RNA species in a single experiment.

• In each round, target RNA molecules are hybridized with FISH probes carrying a single fluorophore (e.g., red, blue, or green).

• The sample is imaged to capture the fluorescent signals.

• DNase I is used to digest the FISH probes, removing them from the sample.

• The same RNA molecules are then hybridized with a new set of probes, now labeled with different fluorophores. This process repeats iteratively over multiple rounds, generating a unique temporal barcode for each RNA species.

Barcoding Example:

To see how seqFISH generates a vast number of barcodes for thousands of genes, let us consider a temporal barcoding example.

The number of uniquely encoded RNA species follows an exponential scaling law:

Total Barcodes = F^N

where:

• F is the number of fluorophores,

• N is the number of hybridization rounds.

For example, if we use 3 fluorophores (red, blue, green) and perform 2 hybridization rounds, we can encode 3² = 9 unique RNA species with the following barcode combinations:

• RR (Red in round 1, Red in round 2)

• BR (Blue in round 1, Red in round 2)

• GR (Green in round 1, Red in round 2)

• RB (Red in round 1, Blue in round 2)

• BB (Blue in round 1, Blue in round 2)

• GB (Green in round 1, Blue in round 2)

• RG (Red in round 1, Green in round 2)

• BG (Blue in round 1, Green in round 2)

• GG (Green in round 1, Green in round 2)

Scaling this further, with 4 fluorophores and 8 hybridization rounds, we can encode 4⁸ = 65,536 RNA species, covering the entire transcriptome.

seqFISH+ (2019)

seqFISH+ builds on seqFISH by introducing an improved probe design and workflow that significantly increases multiplexing capacity and imaging resolution.

How does seqFISH+ probe design increase its multiplexing capacity?

Each primary probe in seqFISH+ consists of:

1. mRNA-complementary sequence (28–30 nucleotides):

   • Targets specific regions of the mRNA for precise hybridization.

   • Designed to avoid off-target binding by optimizing GC content and sequence specificity.

2. Barcode regions (4 regions, each 20 nucleotides):

   • Serve as binding sites for fluorescently labeled readout probes during imaging.

   • Each barcode region corresponds to one barcoding round.

3. Spacer sequences (2 nucleotides):

   • Separate barcode regions to prevent structural interference during hybridization.

4. Flanking primer sequences:

   • Used for amplification during probe synthesis.

Each gene is targeted by 24 primary probes, which ensures robust signal amplification and redundancy for accurate detection.

The structure of each barcode can look like the following:

5’ Primer – Barcode Region R1 (20nt) – 2nt spacer – Barcode Region R2 (20nt) – 2nt spacer – Barcode Region R3 (20nt) – 2nt spacer – Barcode Region R4 (20nt) – 3’ Primer

Barcoding Strategy

• Imaging is performed using three fluorescent channels (e.g., Alexa Fluor 488, Cy3b, and Alexa Fluor 647), with each channel capable of encoding up to 8,000 genes.

• Within each channel, the barcoding process consists of four rounds. Each barcoding round includes 20 hybridization cycles, where readout probes bind to the overhang sequences of the primary probes. These hybridization events generate a unique identifier between 1 and 20, referred to as a ‘pseudocolor’—a numerical code that represents the hybridization state rather than a physical color. In other words, n (n=20 for example) pseudocolours can be thought of as n-bit barcodes in which only one bit contains a ‘1’.

• In simple terms, each barcoding round produces one pseudocolor. With three barcoding rounds, the number of possible barcodes is 20 × 20 × 20 = 8,000, allowing for the detection of 8,000 unique genes per channel. The authors implemented a fourth barcoding round, primarily for error correction.

• Since each fluorescent channel can encode 8,000 genes, using three channels enables the detection of 24,000 genes in total.

Since imaging time is the primary bottleneck in spatial transcriptomics, seqFISH+ achieves an eightfold speed increase compared to seqFISH with expansion microscopy. In seqFISH, detecting 16,384 genes requires four fluorescent channels, eight barcoding rounds, and a 60-fold sample expansion, resulting in 1,920 images per field of view (FOV). In contrast, seqFISH+ encodes genes using 60 pseudocolors across four barcoding rounds, reducing the imaging requirement to 240 images per FOV while detecting 24,000 genes—dramatically cutting imaging time. Additionally, the use of more pseudocolors and a shorter four-round barcode helps minimize accumulated errors over multiple barcoding cycles.

MERFISH (2015): Multiplexed Error-Robust Fluorescence In Situ Hybridization

Multiplexed Error-Robust Fluorescence In Situ Hybridization (MERFISH) is a powerful spatial transcriptomics technique that enables the detection of thousands of RNA species at single-molecule resolution. It builds on smFISH but overcomes its limitations by using a combinatorial binary barcoding system with built-in error correction, allowing for highly multiplexed and accurate RNA imaging.

The core idea of MERFISH: Instead of giving each gene a unique fluorophore, it assigns each RNA molecule a binary barcode (sequence of zeros and ones) that is read out over multiple imaging cycles.

In the first imaging round, fluorescent probes are introduced to tissue samples. Any RNA molecules that bind with the probe are assigned the barcode digit 1 for “ON”. Unbound mRNAs are assigned digit 0 for “OFF”. After imaging, the probes are removed and replaced with a new set for the next hybridization round. This process repeats for N imaging rounds, with each round contributing one bit to the barcode. At the end of the experiment, the fluorescence patterns across all rounds are compiled to reconstruct the binary barcode for each RNA molecule. Computational algorithms match the observed barcodes to a predefined gene library, identifying thousands of RNA species with high accuracy.

Barcoding Example

If an experiment uses 4 imaging rounds, each RNA species gets a 4-bit barcode, like:

Using this approach, with 16 imaging rounds, MERFISH can encode up to 2¹⁶ = 65,536 unique RNA species. With built-in error correction, it can robustly detect 10,000–20,000 genes per experiment, far beyond smFISH or even seqFISH.

Methods by Properties

Based on the original in situ sequencing, smFISH, seqFISH(+), and MERFISH technologies, 9 products have been commercialized on the market.

Imaging-based spatial transcriptomics technologies provide subcellular resolution and enable highly multiplexed RNA detection directly within preserved tissue sections. While most platforms support FFPE and fresh frozen tissues, they differ in customizability, multiomic capabilities, throughput, specificity, and sensitivity, influencing their suitability for different research applications.

Below, we compare major platforms from 10X Genomics, NanoString (Bruker), Vizgen, Resolve Biosciences, Acuity Spatial, Biotechne, Spatial Genomics, Stellaromics, and RebusBio.

Gene Panels: Customizability & Scale

The number of genes each platform can profile varies significantly.

• High-Plex, Customizable Panels:

  • seqFISH+ (Spatial Genomics): 100-1,000 genes

  • MERFISH Ultra (Vizgen): 100-1,000 genes

  • Stellaromics: Demonstrated >1,000 genes in intact mouse brain sample with STARmap PLUS.

  • Rebus Esper with EEL FISH: 1,230 genes

  • 10X Genomics Xenium: 480 to 5,000 genes

  • NanoString CosMx (Bruker): Up to 6,000 genes

• Low-plex, customizable, and optimized gene panels for high specificity:

  • Resolve Biosciences Molecular Cartography: ~100 genes

  • Biotechne RNAscope: ~12 genes

  • Rebus Esper with smFISH: Up to 30 genes

Platforms with customizable panels allow for discovery-based studies where researchers need to detect thousands of genes across tissue sections. Fixed panels, on the other hand, are optimized for high sensitivity and specificity, making them ideal for biomarker validation and targeted studies.

Throughput – Imaging Area Per Week

*Note that data was not provided in other company websites and therefore not included in the graph. Please reach out to individual companies to understand their throughput.

In spatial transcriptomics, throughput—the capture-able image area per week—varies significantly across platforms, each offering distinct advantages and trade-offs. NanoString's CosMx leads with a throughput of 7,800 mm² per week, making it ideal for large-scale tissue studies where extensive spatial coverage is essential. However, higher throughput often comes at the expense of detection sensitivity.

Conversely, platforms like Resolve Biosciences' Molecular Cartography, with a throughput of 52 mm² per week, prioritize high spatial resolution and specificity, enabling detailed analysis of gene expression at the single-cell level. The trade-off here is a reduced capacity to process large tissue areas within the same timeframe.

10X Genomics' Xenium and Vizgen's MERFISH Ultra offer intermediate throughputs of 1,400 mm² and 900 mm² per week, respectively, balancing spatial coverage with resolution and sensitivity. These platforms are suitable for studies requiring a moderate tissue area analysis without compromising on cellular detail.

Tissue Types Profiled: Beyond FFPE & Fresh Frozen

Most FISH-based technologies work with FFPE and fresh frozen tissues, ensuring compatibility with archival pathology samples and clinical studies. However, some platforms support additional sample types:

One distinctive feature of Vizgen’s Merscope and MERFISH Ultra is its capability to profile cultured cells, enabling the examination of spatial intracellular signaling in complex cell cultures such as 3D structures or organoids. Similarly, Stellaromics’ Pyxa platform allows detecting 1,000 transcripts on intact 100-200 micron 3D tissue structure. These capabilities are critical for researchers studying intracellular signaling, stem cell differentiation, and developmental biology.

Multiomic Capabilities: RNA & Protein Profiling

Some platforms enable simultaneous RNA and protein detection, while others focus solely on RNA. Current platforms that provide co-detection include:

• NanoString CosMx (Bruker) (Detection in a sequential workflow) – Up to 64 proteins

• MERFISH Ultra (Vizgen) (Simultaneous RNA and protein profiling on the same tissue section) – Up to 6 proteins

• Biotechne RNAscope Multiomic LS assay – Up to 6 proteins

• Spatial Genomics GenePS – Offers sequential immunofluorescence for protein detection, but number of proteins unspecified.

Multiomic capabilities are valuable in immuno-oncology, tumor microenvironment studies, and spatial proteomics, where linking gene expression to protein localization is crucial.

Sensitivity

Hartman and Satija (2024) conducted a comparative benchmarking analysis of six multiplexed in situ gene expression profiling technologies using publicly available mouse brain datasets.

Among the technologies analyzed, Vizgen’s MERSCOPE exhibited the best balance of sensitivity and specificity, both in its published output and after re-segmentation. MERSCOPE achieved this performance with a large panel of 483 genes, demonstrating that high specificity does not necessarily come at the expense of panel size. 10x Genomics’ Xenium also performed well but used more lenient segmentation boundaries, leading to a higher number of detected molecules per cell but also a significant increase in incorrect molecular assignments.

Resolve Biosciences’ Molecular Cartography and MERFISH showed similar sensitivity, despite having vastly different panel sizes. Molecular Cartography, with only 99 genes, demonstrated strong per-gene sensitivity, whereas MERFISH, with the largest panel (1147 genes), showed that high sensitivity can be maintained even with high plexity. EEL FISH, in contrast, had the lowest sensitivity of all platforms, a well-known tradeoff attributed to its design, which prioritizes fast imaging and reduced autofluorescence by electrophoretically transferring mRNA transcripts to a secondary slide.

Another study by Wang et. al (2023) compared the performance of various in situ imaging technologies in FFPE tissues and tumor microarrays (TMAs). The authors discovered that Xenium had the highest sensitivity across tissues. CosMx and MERSCOPE both detected fewer transcripts than Xenium. MERSCOPE outperformed CosMx in the higher quality (as judged by relative performance across tumor TMA but underperformed it in lower quality normal tissue TMA.

Focusing on performances in FFPE tissue, another study by Cook et. al (2023). They generated spatial transcriptomics data from serial sections of prostate adenocarcinoma using the 10x Genomics Xenium and NanoString CosMx SMI platforms. Additionally, orthogonal single-nucleus RNA sequencing (snRNA-seq) was performed on the same FFPE tissue to establish a reference for the tumor’s transcriptional profiles. The author concluded that although fewer genes are measured (CosMx: 960, Xenium: 377, with an overlap of 125), Xenium consistently demonstrates higher sensitivity, a broader dynamic range, and better alignment with single-cell reference profiles. Conversely, CosMx’s out-of-the-box segmentation outperformed Xenium’s, resulting in noticeable transcript misassignment in Xenium within certain tissue areas. However, the impact of this on the cells’ transcriptional profile was minimal.

Specificity: Accuracy of Gene Detection

Specificity refers to how well a platform correctly identifies closely related RNA sequences while minimizing off-target signals. The study by Wang et. al (2023) demonstrated that Xenium and Merscope demonstrate superior specificity compared to CosMx in FFPE tissues for tumor microenvironments, while CosMx enables detection of larger gene panels.

For applications requiring single-nucleotide specificity (e.g., distinguishing splice isoforms or SNP variants), platforms with error correction and high specificity are better options.

Species

If working with non-human or non-mouse samples, viable options are Vizgen’s MERFISH Ultra, Resolve Biosciences’ Molecular Cartography, and Rebus Esper. While many companies offer custom gene panel design, the cost can be significant due to the extensive optimization required to ensure panel performance.

Future developments

Higher plexity towards whole transcriptome

Today, no imaging-based method can unbiasedly profile the whole transcriptome with high sensitivity. While multimodal measurements are possible, further advances in multiplexing across all modalities could enable spatially resolved, genome-wide measurements at cellular resolution. This, in turn, would allow the generation of single-cell phenotype atlases with spatial context from a single experiment.

Leading this trend, Nanostring announced at AGBT 2025 that its Human Whole Transcriptome Panel will launch in April, with commercial shipments expected this summer. A Mouse Whole Transcriptome Panel is slated for release later in 2025.

Live-cell imaging

Most existing methods focus on fixed cells, limiting the ability to capture dynamic gene expression changes in live-cell contexts. Advances in live-cell imaging and real-time transcriptomic profiling will be important to fully understanding cellular behavior in native environments.

3D atlas construction

Additionally, as the vast majority of tissues are three-dimensional structures, improving spatial profiling methods, whether through serial 2D sections or true 3D imaging, will enable the construction of comprehensive 3D atlases of cells and tissues. However, integrating these datasets across multiple modalities and length scales requires novel computational methods capable of harmonizing high-dimensional spatial and molecular data.

Unified ecosystem of verified computational tools

Standardization in computational pipelines remains a major challenge. Currently, no unified ecosystem exists for imaging-based spatial transcriptomics across the dozens of platforms available. While instrument-provided image segmentation software represents a step forward, it is insufficient for extracting meaningful biological insights. Scientists may want to re-analyze terabytes of imaging data alongside gene expression profiles but lack the necessary computational infrastructure without sophisticated cloud setups.

Many critical questions often arise downstream: How do ligand-receptor interactions vary across different tissue microenvironments? Can users dynamically select and refine spatial clusters for downstream analysis? How can spatial and transcriptomic data be integrated with other omics layers to generate a more complete picture of cellular function? One cannot underestimate the necessity of intuitive and well-designed user interfaces that empower scientists to interact with and interpret their data easily. Thoughtful curation of these tools, combined with an emphasis on accessibility and delightful user experience, will be key to unlocking deep insights from large-scale spatial projects.