Visium HD vs Xenium vs CosMx vs GeoMx vs MERSCOPE vs Stereo-seq

Why platform choice matters in 2025

If you are planning a spatial transcriptomics study in 2025, you are probably asking the same question many PIs and bioinformatics leads ask us first: "Which spatial transcriptomics platform should we actually use?"

That choice is no longer a simple vendor preference. Your spatial transcriptomics platform affects:

• The smallest structures you can resolve (cell, subcellular, or multi-cell spots)
• Whether you can use precious FFPE blocks or only fresh-frozen tissue
• How many genes you can profile per section
• The cost per sample and per project
• The complexity of downstream spatial transcriptomics data analysis

A good platform decision gives you interpretable maps and clear biological stories. A bad one gives you pretty images that reviewers question, or datasets that your team struggles to analyse within realistic timelines.

In this spatial transcriptomics platform comparison, we focus on the commercial systems most often shortlisted in real projects: 10x Genomics Visium / Visium HD, NanoString Xenium, NanoString CosMx SMI, NanoString GeoMx DSP, Vizgen MERSCOPE, and BGI/MGI Stereo-seq. The goal is not to crown a "winner", but to help you choose a spatial transcriptomics platform that fits your tissue type, study design, and budget.

Sequencing vs imaging spatial platforms

Before comparing individual products, it helps to understand the two main technology families used in spatial genomics services today.

Resolution differences between Visium, Xenium and Visium HD on the same ovarian tumor section. (Pérez-Peña E. et al. (2025) Preprints.org)

Sequencing-based spatial transcriptomics

Sequencing-based spatial platforms, suchas Visium HD and Stereo-seq, embed barcodes directly on the slide. Each feature (spot, bin, or nanoball cluster) captures RNA from the tissue above it. After library prep and next-generation sequencing, you obtain a matrix of gene expression values linked to spatial coordinates.

Typical characteristics are:

• Readout: NGS run on an Illumina or compatible instrument
• Gene coverage: whole-transcriptome or very broad capture
• Tissue coverage: relatively large areas per slide
• Resolution: from multi-cell spots down to near-cellular or subcellular bins, depending on platform

These platforms are often the first choice for discovery-oriented spatial transcriptomics sequencing services, tissue atlases, and exploratory oncology projects where you want an unbiased look at the transcriptome.

Spatial indexing-based transcriptomics technologies used by commercial platforms. (Jin Y. et al. (2024) Molecular Cancer)

Imaging-based spatial transcriptomics

Imaging-based platforms, such as Xenium, CosMx, GeoMx, and MERSCOPE, use fluorescence imaging of labelled probes or rolling-circle products. They detect RNA molecules directly in the tissue, often across multiple cycles.

Their typical features include:

• Readout: high-resolution microscopy and image analysis
• Gene coverage: targeted panels, from a few hundred to several thousand genes
• Resolution: single-cell to subcellular localisation
• Sample types: strong support for FFPE and fresh-frozen tissue, depending on system

These technologies are attractive when your questions focus on cell–cell interactions, immune niches, or precise localisation of a curated gene panel rather than global transcriptome coverage. They also pair well with spatial multi-omics analysis that combines RNA and protein markers in the same section.

Key trade-offs for platform selection

In practice, teams usually balance:

• Resolution vs area – tiny pixels vs whole-slide coverage
• Whole-transcriptome vs targeted panels – discovery vs focused validation
• FFPE compatibility – critical for clinical archives and translational studies
• Throughput and cost per case – especially in large cohorts
• Bioinformatics complexity – image-heavy vs sequencing-heavy pipelines

The rest of this guide translates those trade-offs into concrete decisions between Visium HD vs Xenium, CosMx vs GeoMx, MERSCOPE vs Stereo-seq, and other realistic combinations.

At-a-glance spatial transcriptomics platform comparison

A single table can rarely capture every nuance, but it helps to start with the big picture. When we help teams shortlist platforms for a new spatial genomics service project, we usually bin these systems by a few practical dimensions.

Summary table: six popular platforms in 2025

When you draft your internal comparison, consider columns such as:

• Technology type: sequencing-based vs imaging-based
• Resolution: spot-level, single-cell, or subcellular
• Gene throughput: whole-transcriptome vs targeted panel size
• Sample compatibility: FFPE, fresh-frozen, or both
• Tissue area per run: approximate square millimetres or slide fraction
• Typical cost tier: qualitative (entry, mid, premium)
• Bioinformatics load: primarily NGS-centric vs heavy image analysis

Even without exact numbers, organising the six platforms this way makes their roles clear. Visium HD and Stereo-seq cluster in the "broad, transcriptome-wide" group. Xenium, CosMx, MERSCOPE, and GeoMx cover the "targeted, higher-resolution" space with different balances of area, resolution, and panel depth.

Grouping by use case, not brand

Most projects fall into one of a few categories:

• Broad discovery across tissue – map major cell populations and gradients
• Fine-grained niche analysis – zoom into immune niches, tumour margins, or germinal centres
• Large FFPE cohort profiling – many samples with moderate depth
• Atlas-style mapping – very large areas or whole organs

Sequencing-based platforms usually dominate the first and last categories. Imaging systems and ROI-based approaches often anchor the second and third. Thinking in use-case "buckets" makes platform choice much easier than comparing spec sheets line by line.

Quick recommendations by project type

At a high level (and without binding anyone to a single answer):

• New discovery project in fresh-frozen tissue
  • Start with Visium HD or a similar sequencing-based spatial transcriptomics service.
• FFPE oncology cohort with tens to hundreds of samples
  • Consider GeoMx DSP or Visium FFPE for broad profiling, then Xenium or CosMx for targeted follow-up.
• Immune niche or cell–cell interaction study
  • Imaging-based platforms like Xenium, CosMx, or MERSCOPE shine here.
• Brain or developmental atlas with large sections
  • Stereo-seq or high-coverage sequencing approaches can be more scalable.

The next sections unpack why.

Platform profiles and use cases

Here we focus on practical experience: where each platform tends to work well, where it struggles, and what kind of spatial transcriptomics sequencing service or bioinformatics pipeline it typically requires.

10x Genomics Visium & Visium HD

Visium and Visium HD are often the first platforms researchers encounter. They are sequencing-based, with features on the slide capturing RNA from overlying tissue.

• Strengths
  • Whole-transcriptome coverage for discovery work
  • Flexible for many tissues (tumour, brain, developmental, immune)
  • Established workflows and community tools ease spatial transcriptomics data analysis
• Limitations
  • Classical Visium uses multi-cell spots; HD improves resolution but still aggregates several cells in some contexts
  • FFPE performance depends strongly on block quality and pre-analytical handling
• Best-fit projects
  • Early-phase discovery in oncology and neuroscience
  • Pilot projects where you want to explore the full transcriptome before committing to a targeted panel

In our experience, teams that plan to run large cohorts often start with a modest Visium HD pilot, refine regions and markers, and later add a targeted imaging platform for higher-resolution follow-up.

NanoString Xenium

Xenium is an imaging-based platform designed for high-plex spatial panels in both FFPE and fresh-frozen tissue. It uses in situ detection of specific transcripts with high spatial precision.

The classic long-tail question is "Visium HD vs Xenium: which is better for my FFPE cohort?" The answer usually depends on whether you still need whole-transcriptome data.

• Strengths
  • Strong FFPE performance when pre-analytics are controlled
  • High-plex panels tailored for immuno-oncology, fibrosis, and other translational themes
  • Single-cell resolution and clear alignment with pathology workflows
• Limitations
  • Panel design requires thought; broad transcriptome discovery must come from elsewhere
  • Data volumes and image analysis can be heavy for very large panels and tissue areas
• Best-fit projects
  • FFPE-rich tumour cohorts where you already know key pathways and markers
  • Validation of gene signatures first seen in bulk RNA-seq, single-cell, or Visium data

NanoString CosMx SMI

CosMx SMI extends imaging-based spatial transcriptomics down to subcellular resolution with high-plex panels. It is often evaluated alongside Xenium and MERSCOPE.

The "CosMx vs GeoMx" comparison comes up when teams face the trade-off between high-resolution imaging and ROI-based profiling across many samples.

• Strengths
  • Subcellular localisation within each cell
  • Very high gene counts for targeted panels
  • Strong for complex microenvironments such as tumour–immune interfaces
• Limitations
  • Data sets can be extremely large; you need a serious bioinformatics analysis service or in-house pipeline
  • Full-slide coverage at maximum resolution is often impractical; many designs pick ROIs or downsample
• Best-fit projects
  • Detailed dissection of immune synapses, stromal niches, or rare cell states
  • Follow-up studies where a smaller sample set justifies a deeper dive per section

NanoString GeoMx DSP

GeoMx DSP uses a region-of-interest (ROI) model to profile RNA and/or protein across selected tissue regions. It sits between whole-slide platforms and ultra-high-resolution systems.

• Strengths
  • Efficient for FFPE clinical cohorts with many samples
  • Flexible ROI selection guided by pathology and IHC
  • Built-in capabilities for paired RNA–protein spatial profiling
• Limitations
  • Not a true single-cell or subcellular platform; each ROI aggregates multiple cells
  • ROI design is crucial and requires cross-talk between pathologists and molecular teams
• Best-fit projects
  • Biomarker discovery across large FFPE cohorts
  • Studies where you want many patients and fewer sections per case, guided by histology

Vizgen MERSCOPE

MERSCOPE is a multiplexed imaging platform that maps hundreds to thousands of genes at single-molecule resolution in tissue. It is popular in brain and immune system studies.

In "MERSCOPE vs Stereo-seq" discussions, the contrast is often between rich imaging within moderate areas and sequencing-based coverage over very large regions.

• Strengths
  • High spatial precision, particularly in densely packed tissues
  • Good for co-expression patterns, small structures, and layered architectures
  • Mature analysis tools for certain organ systems
• Limitations
  • Panel design and imaging time can limit throughput
  • Scaling to very large tissue areas within a single experiment can be demanding
• Best-fit projects
  • Brain regions, lymphoid organs, or mucosal tissues with complex layer structures
  • Mechanistic studies where spatial patterns matter more than cohort size

BGI/MGI Stereo-seq

Stereo-seq uses DNA nanoball arrays to achieve very high spatial resolution with sequencing-based readout. Bins can be small enough to approximate or even exceed single-cell detail across large tissue areas.

• Strengths
  • High coverage across large sections or even whole embryonic structures
  • Well suited to atlas-style projects and developmental biology
  • Integration with existing NGS lab infrastructure
• Limitations
  • Data volumes are substantial, and pipelines are evolving
  • Access and logistics may depend on regional partners and platforms
• Best-fit projects
  • Organ and developmental atlases
  • Large-area mapping where you still want near-cellular resolution

How to choose a platform for your study

Once you understand the landscape, you still need to decide what to actually run in your next spatial transcriptomics sequencing project.

Key decision criteria: tissue, FFPE status, sample amount

A simple checklist helps most teams narrow down options:

• Tissue type

  Brain, solid tumour, lymphoid tissue, fibrotic organ, or something more unusual? Some platforms have more public datasets and community experience in specific tissues.

• Preservation method

  Are your samples FFPE, fresh-frozen, or both? Can you collect prospective fresh-frozen material, or are you tied to archived blocks?

• Amount of material

  Do you have multiple sections per case, or just one slide per patient? This determines whether pilot experiments and repeated runs are feasible.

• Number of samples

  Is this a 10-sample mechanistic study or a 200-sample translational cohort? Some platforms scale more gracefully than others.

Budget scenarios: pilots vs scaled cohorts

In real projects, budgets drive strategy as much as biology:

• High-resolution pilot, smaller N
  • Option: Xenium, CosMx, MERSCOPE, or Stereo-seq on 10–20 representative samples
  • Goal: understand microenvironments, refine gene sets, test feasibility
• Moderate-resolution cohort, larger N
  • Option: Visium FFPE/HD or GeoMx on 50–200+ cases
  • Goal: capture robust spatial patterns and biomarkers across many patients

A hybrid approach is common: start with Visium or Stereo-seq for broad mapping, then switch to Xenium or CosMx for a specific niche in a subset of samples. This also aligns well with integrated epigenomic sequencing and bioinformatics analysis services, where you might combine chromatin accessibility, methylation, and spatial RNA data.

Study goals: microenvironment, atlas, or biomarker validation?

Clarifying your primary goal helps avoid over-engineering the experiment:

• Tumour microenvironment and immune niches
  • Imaging-based systems plus single-cell RNA-seq reference data are often preferred.
• Atlas-style mapping of an organ or developmental stage
  • High-coverage sequencing (Visium HD, Stereo-seq) fits better, especially with large areas.
• Biomarker validation in FFPE archives
  • FFPE-friendly platforms (GeoMx, Xenium, Visium FFPE) are the practical options.

FAQ: common questions when choosing a spatial platform

Below are some of the questions we hear frequently in discussions and see often in AI tools.

Q1. Visium HD vs Xenium: which is better for my FFPE cohort?

If you still need whole-transcriptome data to discover pathways or cell states, Visium HD (or similar sequencing-based FFPE workflows) are often the first step. If you already have a gene list from previous RNA-seq or single-cell work and want pathology-aligned, high-plex panels at single-cell resolution, Xenium becomes more attractive. Many teams run Visium HD on a pilot set, then follow up selected regions with Xenium.

Q2. CosMx vs GeoMx: when does subcellular resolution really matter?

CosMx provides subcellular resolution and very high-plex panels, but with heavier imaging and analysis demands. GeoMx profiles ROIs with lower resolution but can scale across more samples at lower per-case cost. If your questions hinge on cell–cell contacts, synapses, or subcellular localisation, CosMx is worth the overhead. If you are screening many FFPE cases for pathway-level signals, GeoMx is often the more pragmatic first choice.

Q3. MERSCOPE vs Stereo-seq: which fits brain and atlas projects?

MERSCOPE is strong for detailed imaging in specific brain regions, with rich in situ visualisation. Stereo-seq is advantageous when you need near-cellular resolution over very large areas, such as whole sections or developmental structures. Your choice depends on whether you prioritise image-centric analysis or sequencing-centric pipelines and how large each section is.

Q4. Is spatial transcriptomics always better than single-cell RNA-seq?

No. Single-cell RNA-seq still offers unmatched flexibility for profiling many samples and conditions without spatial constraints. Spatial transcriptomics is most valuable when tissue architecture, neighbourhoods, or microenvironments are central to the biology. Many successful programmes combine single-cell and spatial data rather than replacing one with the other.

Q5. How many samples do I need for a spatial cohort?

There is no universal number. For exploratory mechanistic work, 5–10 high-quality cases per group can already reveal spatial patterns, especially when combined with orthogonal data. For translational projects aiming at robust biomarkers, you typically need dozens of cases per group and a design aligned with your downstream statistics. A dedicated bioinformatics analysis service can help simulate power under realistic assumptions.

Lessons from real spatial projects

Even with a good platform choice, practical details often determine whether a dataset becomes a figure or a troubleshooting story.

Sample handling makes or breaks data quality

From a practical standpoint, pre-analytical factors are as important as spec sheets:

• Tissue should be processed quickly and consistently after collection.
• For fresh-frozen samples, embedding, freezing rate, and storage time affect RNA integrity and morphology.
• For FFPE, fixation duration, block age, and sectioning quality all influence signal.

Teams that invest early in standard operating procedures for embedding, sectioning, and slide handling usually see fewer failures, regardless of platform.

Panel design and pilot strategies

For panel-based platforms, careful design prevents disappointment:

• Start from real data (bulk, single-cell, or previous spatial runs) rather than intuition alone.
• Allocate panel capacity to a mix of housekeeping, known markers, pathway genes, and exploratory candidates.
• Use pilot sections to confirm that key markers are expressed at detectable levels and that spatial patterns align with existing knowledge.

Treat the first run as a calibration, not as the definitive answer. This mindset reduces pressure and allows you to refine panels before scaling.

Bioinformatics bottlenecks and realistic timelines

Spatial datasets stress both compute and people:

• Imaging-based runs generate large tiled images and segmentation masks.
• Sequencing-based runs generate many millions of reads per section.
• Integration with single-cell data adds another layer of complexity.

When planning your spatial transcriptomics data analysis service or internal pipeline, account for:

• Quality control and filtering
• Segmentation and cell calling (for imaging systems)
• Normalisation, clustering, and annotation
• Spatial statistics and neighbourhood analyses
• Integration with other omics, including epigenomic sequencing

Most teams underestimate analysis time the first time they run a spatial project. Building realistic timelines based on prior experience helps keep expectations aligned.

Case-style patterns we see repeatedly

Although project details vary, some patterns repeat:

• Teams that start with a small, well-designed pilot generate cleaner protocols and more convincing figures later.
• Projects that mix one broad discovery modality (e.g. Visium HD or Stereo-seq) with one targeted imaging modality (e.g. Xenium, CosMx, MERSCOPE) tend to extract more actionable insights.
• Collaborations that align pathology, wet-lab, and bioinformatics early handle ROI selection, QC, and interpretation more smoothly.

These lessons are independent of vendor and reflect how spatial omics behaves in real laboratories.

Applications of spatial transcriptomics in cancer research. (Jin Y. et al. (2024) Molecular Cancer)

From platform choice to project launch

A good spatial study starts long before the first slide goes on an instrument.

Pre-launch checklist for spatial transcriptomics projects

Before you lock in a platform, it helps to have the following on one page:

• Clear primary and secondary biological questions
• Tissue types, preservation methods, and availability per case
• Expected sample numbers and budget range
• Preferred platforms or a short shortlist
• Plan for bioinformatics analysis and data sharing

Walking into a vendor discussion or a spatial transcriptomics sequencing service consultation with this information speeds up scoping and avoids surprises.

How we support platform selection, study design, and analysis

An experienced partner does more than just run a chosen assay. In a typical engagement, we:

• Review your study aims, tissue constraints, and existing omics data
• Provide an unbiased spatial transcriptomics platform comparison tailored to your case
• Co-design pilots and main cohorts, including slides per case and ROIs
• Define a bioinformatics plan that covers raw data processing, spatial statistics, and integration with other assays, including epigenomic sequencing and standard RNA-seq

This integrated view reduces the risk of fragmented datasets that are difficult to interpret across technologies.

Talk to us about your spatial platform shortlist

If you already have a shortlist—perhaps Visium HD vs Xenium, CosMx vs GeoMx, or MERSCOPE vs Stereo-seq—the next step is turning that list into a concrete experimental plan.

Share a short summary of your project, including tissue type, sample numbers, and key questions. Our team can help you refine platform choice, design a realistic pilot, and align spatial transcriptomics sequencing and bioinformatics analysis services with your wider multi-omics strategy.

A good decision at this stage will save you time, budget, and reviewer questions later—and give you spatial maps that genuinely move your research forward.

References

1. Lim, H.J. et al. A practical guide for choosing an optimal spatial transcriptomics technology from seven major commercially available options. BMC Genomics 26, 47 (2025).
2. Chen, T.-Y., You, L., Hardillo, J.A.U. & Chien, M.-P. Spatial transcriptomic technologies. Cells 12, 2042 (2023).
3. Du, J. et al. Advances in spatial transcriptomics and related data analysis strategies. Journal of Translational Medicine 21, 330 (2023).
4. Jin, Y. et al. Advances in spatial transcriptomics and its applications in cancer research. Molecular Cancer 23, 129 (2024).
5. Pérez-Peña, E. & colleagues. A practical guide to spatial transcriptomics: lessons from over 1000 samples. Preprints (2025).
6. General practical review (Valihrach et al., Mol. Aspects Med.) Valihrach, L., Zucha, D., Abaffy, P. & Kubista, M. A practical guide to spatial transcriptomics. Molecular Aspects of Medicine 97, 101276 (2024).