Which Single-Cell Platform Is for What? A 2026 Field Guide

A deep dive into experimental, spatial, multi-omic, and computational platforms — and how to actually pick the right one

If you’ve planned a single-cell experiment in the last year, you know the feeling. You sit down to draft a methods section and realize you have to choose between droplet capture, combinatorial indexing, microwells, probe hybridization, fixed cells, fresh cells, FFPE, multiome, CITE-seq, TEA-seq, Visium, Xenium, CosMx, Stereo-seq, Scanpy, Seurat, scvi-tools — and you’re not sure where to start.

This article is the guide I wish someone had handed me.

The single-cell genomics platform landscape: five interconnected domains.

The single-cell field has reached a strange crossroads in 2026. We have more powerful technologies than ever, more benchmarking papers than ever, and more confused PIs than ever. The global single-cell market hit roughly $1.7 billion in 2025 and is still growing at over 13% per year (DeciBio, 2024) — and every quarter brings a new kit, a new chemistry, a new “game-changing” platform.

Here’s the thing: no single platform wins at everything. The right choice depends on your biological question, your samples, your budget, and your downstream analysis plan. This guide breaks down the landscape into six domains and ends with a decision framework you can actually use on Monday morning.

Part I — Experimental scRNA-seq Platforms

The experimental platforms split into four families based on how they isolate single cells: droplet emulsions, microwells, combinatorial indexing, and plate-based. A 2026 benchmarking study from Fred Hutch (Gratz et al., NAR Genomics and Bioinformatics) put ten commercial kits head-to-head on PBMCs from a single donor, giving us the cleanest comparison to date.

Throughput and sensitivity comparison across major scRNA-seq platforms.

🔵 10x Genomics Chromium — The Default Choice

10x Chromium is still the platform most labs reach for first, and for good reason. The ecosystem is enormous, the protocols are battle-tested, and the data formats are universally supported.

The current product line includes three flavors:

• 3′ Gene Expression (v3.1 / GEM-X) — the workhorse for most discovery experiments
• 5′ Gene Expression (v2) — needed if you want V(D)J immune receptor profiling
• Chromium Flex — uses probe hybridization instead of reverse transcription, works on fixed cells and FFPE, and supports up to 2.56 million cells per run through 16-sample multiplexing

Here’s the surprise from the Gratz et al. benchmark: the Flex kit had the best analytical performance of any platform tested. Probe hybridization just works better than RT, and fixation lets you batch and ship samples without sacrificing quality.

The downside? Flex only supports human and mouse probe sets, you lose the ability to detect genetic variants, and you still need a $60K+ Chromium X instrument.

🟢 Parse Biosciences Evercode — The Scale Champion

Parse has quietly become 10x’s most serious competitor. Their split-pool combinatorial barcoding strategy needs zero specialized hardware — cells go through multiple rounds of in-situ barcoding in standard 96-well plates, and that’s it.

The new Evercode WT PENTA V3 kit (launched February 2025) supports 5 million cells across 384 samples in a single experiment. That’s the highest throughput on the market, full stop.

Three major cell-partitioning strategies: droplet, microwell, and combinatorial indexing

The benchmarking data on Parse vs. 10x is genuinely interesting. In a BMC Genomics study on mouse thymocytes, Parse detected nearly twice the number of genes per cell as 10x — but 10x had lower technical variability and tighter cell-state annotation. Parse is also better at picking up rare cell types like plasmablasts and dendritic cells.

In February 2025, Parse invalidated all of 10x’s patent claims against them at the Patent Trial and Appeal Board. The competitive landscape is officially open.

💡 TL;DR: If you need maximum throughput with no instrument and don’t mind a multi-day workflow, Parse wins. If you need the cleanest single-experiment data with the broadest tool support, 10x wins.

🟡 The Other Players Worth Knowing

BD Rhapsody uses microwells and was the cost-performance winner in the Gratz benchmark. It also supports AbSeq for protein co-detection and full-length TCR/BCR.

Scale Biosciences mirrors Parse’s combinatorial approach but is the only commercial vendor offering single-cell DNA methylation — a unique capability if epigenetic state matters for your question.

Fluent BioSciences PIPseq does droplet capture by vortex mixing — no microfluidics, no instrument. Lower sensitivity than 10x, but the simplest workflow on the market.

Singleron offers something neither 10x nor Parse can match: simultaneous whole-genome plus whole-transcriptome detection from the same cell via their AccuraSCOPE platform.

Part II — Multi-Omic Platforms (When RNA Isn’t Enough)

Once you’re asking questions about gene regulation rather than just gene expression, you need more than RNA. Multi-omic assays measure two or more molecular layers from the same cell.

Modality support matrix across single-cell platforms.

🧬 10x Multiome (RNA + ATAC) — The Regulatory Workhorse

The 10x Epi Multiome captures gene expression and chromatin accessibility from the same nucleus. It’s become the default platform for anyone studying gene regulatory networks, enhancer-gene linkages, or cell-state transitions.

Practically speaking, you get 500–10,000 nuclei per sample, up to 80,000 per run, with recommended depths of 20K read pairs/nucleus for GEX and 25K for ATAC. The downstream analysis stack is rich: SCENIC+ for enhancer-driven regulatory networks, ArchR or Signac for ATAC processing, and Pando or EpiRegulon for transcription factor inference.

The catch: nuclei prep quality is everything. Ambient DNA contamination ruins ATAC libraries, so DNase I treatment is non-negotiable.

🎯 CITE-seq (RNA + Protein)

CITE-seq adds oligo-conjugated antibodies (sold as TotalSeq by BioLegend) to your standard scRNA-seq workflow. Surface protein expression gets quantified alongside the transcriptome, bridging flow cytometry phenotyping with transcriptomic profiling.

It’s particularly powerful in immunology where surface markers define functional states that aren’t always obvious from transcripts alone.

🔺 TEA-seq (RNA + ATAC + Protein) — Trimodal

TEA-seq is the trimodal frontier: Transcripts, Epitopes, and Accessibility from the same cell, built on the 10x Multiome platform. The trick is isotonic permeabilization, which preserves chromatin state while still letting antibodies bind surface epitopes.

🧪 CAT-ATAC — RNA + ATAC + CRISPR

The newest entrant: CAT-ATAC (Cell Reports Methods, 2025) captures CRISPR guide RNAs alongside RNA and ATAC on the 10x Multiome platform. This means you can run a perturbation screen and read out the regulatory consequences — not just expression changes — at single-cell resolution. For drug-resistance and developmental biology work, this is a meaningful step forward.

Part III — Spatial Transcriptomics: The Resolution Wars

Spatial methods restore the dimension that dissociation throws away: where each cell actually lives in the tissue. The field splits cleanly into sequencing-based (sST) and imaging-based (iST) platforms, and a landmark 2025 benchmark from Ren et al. in Nature Communications put four leading platforms head-to-head on serial human tumor sections with CODEX protein validation.

Spatial resolution vs. gene coverage trade-off across platforms.

📊 Sequencing-Based: Whole Transcriptome, Lower Resolution

Visium HD (10x Genomics, 2024) is the discovery default. It captures the whole transcriptome (~18,000 genes) at 2 µm resolution on FFPE tissue. If you want to find new spatial patterns without bias, this is where you start.

Stereo-seq (BGI) offers the highest resolution of any sequencing-based platform — 220 nm spot arrays — and a large 10×10 mm capture area. But there’s a catch the Lim et al. 2025 practical guide calls out: Stereo-seq shows notable lateral diffusion in some tissues, and at single-cell-equivalent bin sizes, you only detect a few hundred genes. Visium HD costs about twice as much per library, but the data is cleaner.

🔬 Imaging-Based: Single-Cell Resolution, Targeted Panels

Xenium 5K (10x Genomics) — uses padlock probes to detect 5,001 genes at subcellular resolution with x, y, z transcript coordinates. No sequencing cost. It handles degraded RNA well and is the most affordable imaging-based option. If you have FFPE tumor blocks and a hypothesis to test, Xenium is usually the right answer.

CosMx 6K (NanoString) — the largest panel of any imaging platform (6,175 genes) and the only one that does simultaneous protein co-detection. If you need both modalities in the same section, CosMx is the choice.

MERSCOPE (Vizgen) — implements MERFISH at ~1,000 genes. The Lim et al. guide notes that MERSCOPE outperforms Xenium with high-quality fresh-frozen RNA, while Xenium wins on degraded samples. Strong neuroscience track record.

GeoMx DSP (NanoString) — the odd one out. It’s not single-cell; it profiles user-defined ROIs at ~20–200 cell resolution with both RNA and protein. Use it when you need deep profiling of specific structures (germinal centers, tumor margins, vascular niches) rather than whole-section maps.

💡 TL;DR: Discovery → Visium HD or Stereo-seq. Hypothesis-testing → Xenium (FFPE-friendly, cheap) or CosMx (largest panel + protein).

Part IV — Computational Frameworks (The Choice That Quietly Shapes Your Results)

Here’s something most papers don’t mention: the package you choose can change your biological conclusions. Rich et al. (2024) showed that Seurat and Scanpy — supposedly running the same workflow — produce noticeably different log fold-changes and adjusted p-values, with thousands of marker genes flipping between the two.

Computational ecosystem map: Scanpy and Seurat at the center, satellite tools by function.

📊Scanpy (Python)

Scanpy is the dominant Python framework, built around the AnnData object. It’s part of the broader scverse ecosystem alongside scvi-tools, Squidpy (spatial), scirpy (immune receptors), and CellRank (fate mapping). Scanpy scales beautifully — it can comfortably handle datasets of millions of cells, and with GPU acceleration via NVTabular, it processes 10 million cells in 12 minutes.

Pick Scanpy if: you work in Python, your datasets are large, you want deep-learning interoperability, or you’re building reproducible pipelines.

📊 Seurat v5 (R)

Seurat from the Satija Lab is the most widely used R toolkit. Its anchoring method is excellent for integration across batches and modalities. Seurat v5 natively handles spatial data, multiome (via Signac), and CITE-seq. The R/Bioconductor ecosystem around it (Monocle 3, CellChat, Signac) is mature and well-documented.

Pick Seurat if: you work in R, you collaborate with statisticians who live in Bioconductor, or you want the smoothest experience for label transfer and reference-based annotation.

📊 scvi-tools (Deep Learning)

scvi-tools brings probabilistic deep learning to single-cell analysis via variational autoencoders. The single-cell best-practices community consistently identifies scVI and scANVI as the top performers for complex integration tasks — multi-batch, multi-tissue, multi-study atlases. The suite extends to totalVI (CITE-seq), PeakVI (ATAC), and scArches (transfer learning to reference atlases).

📊 The Specialist Tool Belt

Beyond the core trio, every workflow pulls from a specialist toolbox:

• Cell Ranger — raw 10x processing (still the gold standard)
• Harmony — fast, lightweight batch correction
• CellBender — ambient RNA removal
• Monocle 3 / scVelo / CellRank — trajectory and dynamics
• SCENIC+ / Pando / EpiRegulon — gene regulatory networks from Multiome
• CellTypist / scGPT — automated annotation (>95% accuracy on immune cells, pretrained on 33M cells)
• Squidpy / Giotto — spatial neighborhood analysis
• CellChat / LIANA+ — cell-cell communication

Part V — Public Data: Where to Actually Find Cells

You don’t always need to generate your own data. The challenge is knowing which resource to use.

Tiered architecture of public single-cell data resources.

🗄️ Tier 1 — Archives (GEO, SRA, ArrayExpress). Use these when you need raw files linked to a specific publication. Indispensable for reproducibility, painful for cross-study exploration.

🌐 Tier 2 — Curated Discovery Portals. This is where most exploratory work should start. CZ CELLxGENE Discover hosts over 1,550 datasets and 169 million cells with standardized metadata as of late 2024 (NAR, 2025). It gives you an Explorer for interactive visualization, a Gene Expression tool for cross-corpus heatmaps, and Census for programmatic access via TileDB-SOMA. Other strong options: Single Cell Expression Atlas (EMBL-EBI), Broad Single Cell Portal.

🎯 Tier 3 — Specialist Atlases. When the question is narrow, the specialist atlas wins. Allen Brain Cell Atlas for neuroscience. TISCH2 for tumor immune microenvironment. Tabula Sapiens for healthy multi-organ reference. HuBMAP for spatial-plus-single-cell tissue maps.

💡 The rule of thumb: Discovery → Tier 2. Specific biological question → Tier 3. Need the original deposited data → Tier 1.

Part VI — The Decision Framework

Here’s the part you actually came for. Walk through these questions in order.

Platform selection decision framework.

Step 1 — What’s your primary biological question?

🔍 Cell-type discovery and heterogeneity → Standard scRNA-seq. 10x Chromium 3′/5′ for the gold standard, Parse Evercode if you need maximum throughput, 10x Flex if you have FFPE.

⚙️ Gene regulation and enhancer-gene networks → 10x Multiome (RNA + ATAC), analyzed with SCENIC+ or Pando. Add TEA-seq if you also need surface protein.

🗺️ Tissue architecture and spatial expression → Discovery: Visium HD or Stereo-seq. Hypothesis-testing: Xenium 5K (cheap, FFPE-friendly) or CosMx 6K (largest panel + protein).

🛡️ Immune receptor profiling → 10x 5′ + V(D)J for the cleanest data, Parse Evercode TCR/BCR for scale, BD Rhapsody for AbSeq + receptors.

🧪 Functional perturbation screens → Perturb-seq on 10x or Parse for transcriptomic readout; CAT-ATAC on 10x Multiome if you also want chromatin response.

Step 3 — What’s your budget reality?

• No instrument budget? → Parse, Scale, Fluent PIPseq (all instrument-free)
• Want to skip sequencing costs? → Imaging spatial: Xenium, CosMx, MERSCOPE
• Tight per-sample budget but need many samples? → BD Rhapsody (cost-performance winner)
• Need million-cell scale on a single experiment? → Parse PENTA V3 (5M) or Scale QuantumScale XL (4M)

What Actually Matters

If you take one thing away from this guide, it’s this: the platform should match the question, not the other way around. Too many experiments get designed around what’s already in the lab, and too many papers report on whichever kit had a representative on campus that quarter.

A few patterns I keep seeing in 2026:

1. Combinatorial indexing is winning the scale war. When you need millions of cells across hundreds of samples, instrument-free platforms (Parse, Scale) are increasingly the obvious choice.
2. Probe-based chemistry is winning on quality. 10x Flex outperformed every RT-based platform in the Gratz benchmark. Probes just work better for fixed and FFPE samples.
3. Spatial is splitting cleanly into discovery vs. testing. Whole-transcriptome platforms (Visium HD, Stereo-seq) for hypothesis generation, targeted imaging platforms (Xenium, CosMx) for hypothesis testing. Pick one for each phase.
4. Multi-omic is becoming the new default for regulation studies. RNA-only is no longer enough if you want to publish a regulatory mechanism. Multiome + SCENIC+ is the new baseline.
5. Computational choice silently shapes biology. Lock your package versions, document everything, and don’t assume Scanpy and Seurat give the same answer.

📚 Key References

• Gratz et al. (2026). Evaluating commercial single-cell RNA sequencing technologies. NAR Genomics and Bioinformatics.
• Ren et al. (2025). Systematic benchmarking of high-throughput subcellular spatial transcriptomics platforms. Nature Communications.
• Lim et al. (2025). A practical guide for choosing an optimal spatial transcriptomics technology. BMC Genomics.
• Filippov et al. (2024). Comparative transcriptomic analyses of thymocytes using 10x and Parse. BMC Genomics.
• Rich et al. (2024). The impact of package selection and versioning on single-cell RNA-seq analysis. bioRxiv.
• CZ CELLxGENE Discover (2025). A single-cell data platform. Nucleic Acids Research.
• Swanson et al. (2021). TEA-seq: trimodal single-cell measurement. eLife.
• González-Blas et al. (2023). SCENIC+: enhancer-driven gene regulatory networks. Nature Methods.

If this guide helped, follow Multiome Academy for more deep dives into single-cell and multiome bioinformatics. We publish tutorials, books (including The Single-Cell Codex and the upcoming Bulk RNA-Seq Codex), and tools to help you go from raw data to biological insight.

Have a platform you think I missed, or disagree with a recommendation? Let me know in the comments — the field moves fast and I’d love to hear what’s working in your lab.