Transcriptomics

Spatial transcriptomics analysis: tissue architecture from Visium spots to spatial domains and cell-type deconvolution

Spatial transcriptomics maps gene expression to tissue coordinates and keeps architecture that dissociated scRNA-seq does not retain [1]. Pepkio runs version-pinned workflows from FASTQs or vendor outputs to spatial domains, deconvolution, and domain-level statistics. Custom inputs, outputs, and methods are scoped at kickoff. For academic, biotech, and pharma clients on 10x Visium CytAssist, typical depth is ≥25,000 read pairs per tissue-covered spot [2], with scripts, figures, and a Methods draft.

Key facts

Key facts about Spatial Transcriptomics
FactValue
Supported platforms / instrumentsPrimary: 10x Genomics Visium v1/v2 (CytAssist), Visium HD, Xenium In Situ (Prime 5K). Illumina or MGI DNBSEQ-sequenced Visium FASTQs when scoped at kickoff. NanoString CosMx SMI, MERFISH, and legacy Spatial Transcriptomics arrays on request
Input requirementsVisium v2: ≥25,000 read pairs/tissue-covered spot, 4,992 spots per 6.5 mm capture area [2]. Visium HD (6.5 mm, 100% tissue): ≥275M read pairs (FFPE), ≥500M (Fixed Frozen), ≥700M (Fresh Frozen) [2]. Xenium: vendor cell-segmented outputs or raw bundles; FFPE or fresh frozen sections
Reference builds supportedHuman GRCh38-2024-A (GENCODE v44 / Ensembl 110); mouse GRCm39-2024-A (GENCODE vM33 / Ensembl 110); Visium FFPE probe sets per 10x reference bundles; custom references on request
Primary tools (with versions)Space Ranger 3.1.3; Scanpy 1.12.1; Squidpy 1.8.1; Seurat 5.2.1; pyBanksy 1.3.4; cell2location 0.1.5; SPOTlight 1.6.3; SpotClean 1.12.1; harmonypy 1.2.3; bin2cell (Visium HD, on request)
Typical turnaround time4-6 weeks (standard Visium cohort, 4-12 sections); 6-10 weeks (Visium HD cell reconstruction, multi-slice integration, or Xenium). Confirmed at kickoff
Deliverable formats.h5ad, .rds, Space Ranger outputs, spatial coordinate files; PDF/SVG spatial plots; HTML QC report; documented R/Python scripts; Methods draft
Key cited best-practice referenceMoses & Pachter [1], Nature Methods; Salas et al. [3], Nature Methods; Long et al. [4], Genome Biology
Custom / bespoke analysisNon-standard inputs, outputs, and methods scoped at kickoff, for example Visium HD bin2cell segmentation, Xenium segmentation QC, cross-platform integration, or client-specified spatial statistics

What is spatial transcriptomics?

Spatial transcriptomics assigns RNA counts to physical locations on a tissue section, producing a spot- or cell-level count matrix linked to (x, y) coordinates and, when available, a registered H&E image. Unlike dissociated scRNA-seq, expression is measured in situ, so cell-cell neighborhoods and tissue domains remain intact [1, 5]. On 10x Visium, 55 µm spots typically capture multiple cells, a mini-bulk readout that needs spatial statistics or deconvolution for cell-type claims [2, 6]. Pepkio starts from FASTQs, Space Ranger outputs, or preprocessed objects and returns annotated spatial datasets with QC at each step. Custom inputs, deliverable formats, and analyses beyond the standard workflow are agreed at kickoff. See the spatial transcriptomics glossary.

When should you use spatial transcriptomics?

Spatial transcriptomics is a fit when the biological question depends on where genes are expressed, such as tumor-stroma boundaries, immune niches, or anatomical layers, not only which cell types are present.

Comparison of spatial transcriptomics, scRNA-seq, and bulk RNA-seq
ApproachBest forLimitationsApproximate cost range
Spatial transcriptomics (Visium, Xenium)Tissue domains, cell-cell niches, spatial DE, microenvironment architectureSpot mixing on Visium; platform-specific capture; sectioning and registration constraintsLibrary prep + sequencing and bioinformatics vary by platform, section count, and depth
scRNA-seq (droplet)Cell-type discovery, rare populations, trajectory inferenceDoes not retain native spatial context; dissociation stress and ambient RNA artifactsLower per-cell sequencing cost than subcellular spatial; no tissue architecture
Bulk RNA-seqCondition contrasts when tissue is homogeneousAverages across cell types and locations; no spatial resolutionLower per-sample cost than spatial for modest cohorts
  • Oral squamous cell carcinoma: Integrative spatial transcriptomic analysis of HPV-negative OSCC identified conserved tumor core and leading-edge architectures associated with survival and targeted therapy response [7].
  • Lung adenocarcinoma progression: Spatial profiling of 30 lung tumors with Visium and Xenium linked tumor-immune interactions at invasive frontiers to adenocarcinoma progression stages [8].
  • Triple-negative breast cancer: Visium across 92 TNBC patients defined nine spatial archetypes, revealing ecosystem heterogeneity with potential clinical relevance [9].

How the analysis works — step by step

  1. 1. Validate inputs and sample metadata

    Pepkio confirms FASTQ pairing, Visium slide serial numbers, CytAssist or H&E image availability, and Xenium bundle completeness when applicable. Platform, preservation method, capture-area coverage, and covariates are recorded in sample_manifest.csv. Sub-threshold depth or incomplete coverage is flagged before downstream work [2, 10].

    Tools and outputs

    Tools used: fastqc / fastp as needed; Loupe Browser tissue-coverage estimates when client provides them

    Output: sample_manifest.csv with library IDs, platform, read counts, capture-area coverage, and QC flags

  2. 2. Align reads and generate spatial count matrices

    For 10x Visium and Visium HD, Pepkio runs spaceranger count with GRCh38-2024-A or GRCm39-2024-A references [2]. Fraction reads in tissue, reads per spot, and genes per spot are compared against vendor ranges. Preprocessed Space Ranger or Xenium outputs can be imported when clients provide them.

    Tools and outputs

    Tools used: Space Ranger 3.1.3

    Output: filtered_feature_bc_matrix/, spatial/, metrics_summary.csv, web_summary.html; Visium HD binned matrices when applicable

  3. 3. Import spatial objects and register images

    Count matrices and coordinates are imported into AnnData or Seurat objects with raw UMI counts in a dedicated layer and H&E or CytAssist images in the spatial slot [11]. Xenium segmented cell matrices retain segmentation QC flags.

    Tools and outputs

    Tools used: Scanpy 1.12.1 (read_visium) or Seurat 5.2.1 (Load10X_Spatial)

    Output: Per-sample .h5ad or .rds with counts layer, spatial coordinates, and image metadata

  4. 4. QC spots or segmented cells

    Spots or cells with low UMI counts, low gene complexity, or high mitochondrial fractions are flagged using sample-adaptive thresholds [1, 3]. For Visium, fraction reads in tissue, sequencing saturation, and spot counts are compared to vendor recommendations [2].

    Tools and outputs

    Tools used: Scanpy 1.12.1 or Seurat 5.2.1

    Output: Filtered object; QC plots for nCount_Spatial, nFeature_Spatial, percent.mt; sample_qc_summary.csv

  5. 5. Correct spot swapping on Visium

    Visium spots can capture transcripts from neighboring locations, a spot-swapping effect that inflates spatial correlation [12]. When raw and background counts are available, Pepkio applies SpotClean before domain calling. Visium HD and Xenium skip this step or use platform-specific QC.

    Tools and outputs

    Tools used: SpotClean 1.12.1

    Output: Decontaminated count matrix; spot_contamination_rate in metadata; SpotClean diagnostic plots overlaid on tissue

  6. 6. Normalize and select spatially informative features

    Library-size normalization and log transformation suppress extreme count values while retaining spatial structure [4]. Pepkio selects highly variable genes for dimensionality reduction and computes spatially variable genes (SVGs) with Squidpy Moran's I, validating top SVGs on the tissue image [3, 11].

    Tools and outputs

    Tools used: Scanpy 1.12.1; Squidpy 1.8.1

    Output: Normalized layers; highly_variable_genes.csv; svg_gene_table.csv with Moran's I scores and FDR

  7. 7. Build spatial graphs and identify tissue domains

    A spatial neighborhood graph connects each spot or cell to its physical neighbors [11]. Pepkio identifies tissue domains with BANKSY, which combines expression and neighborhood context, or Leiden clustering on spatial embeddings, mapping domain labels onto H&E for biological review [4, 13].

    Tools and outputs

    Tools used: Squidpy 1.8.1; pyBanksy 1.3.4; leidenalg 0.10.2

    Output: domain_assignments.csv; spatial domain maps on H&E; neighborhood graph object

  8. 8. Deconvolve cell-type proportions when a reference is available

    Visium spots contain multiple cell types; deconvolution estimates per-spot abundance from a matched scRNA-seq reference [6]. Pepkio trains cell2location or SPOTlight signatures and validates with marker genes [14, 15]. Reference-free deconvolution is scoped when no atlas exists.

    Tools and outputs

    Tools used: cell2location 0.1.5 or SPOTlight 1.6.3

    Output: deconvolution_proportions.csv (spot × cell type); spatial abundance maps; reference signature diagnostics

  9. 9. Integrate batches or multiple sections

    Multi-patient or multi-slide projects are integrated with Harmony when shared spatial domains are present [16]. Cross-slide 3D alignment or Visium-Xenium integration is scoped separately when required [4].

    Tools and outputs

    Tools used: harmonypy 1.2.3 or Seurat 5.2.1 integration workflows

    Output: Harmonized embedding; before/after spatial plots by batch and condition; integration QC metrics

  10. 10. Test differential expression and package deliverables

    Domain-wise or condition-wise DE uses pseudobulk limma-voom with ≥3 replicates per group; spot-level Wilcoxon tests with Benjamini-Hochberg FDR when pseudobulk is not feasible [1]. Results export as ranked gene lists with log₂ fold-change and adjusted p-values.

    Tools and outputs

    Tools used: limma 3.62.2 / edgeR 4.4.2 (R) or Scanpy rank_genes_groups (Python)

    Output: deg_by_domain.csv; spatial feature plots; final .h5ad/.rds; scripts; Methods draft

What Pepkio delivers

Processed data files

  • .h5ad, .rds, Space Ranger outputs, coordinate files, count matrices, and metadata (sample_id, batch, condition, QC metrics, domain, cell_type when deconvolved).

Figures (PDF/SVG)

  • Spot QC plots, SpotClean maps, spatial feature plots, domain maps on H&E, neighborhood enrichment heatmaps, deconvolution maps, DE volcano plots.

Tables

  • spot_metadata.csv, sample_qc_summary.csv, domain_assignments.csv, svg_gene_table.csv, deconvolution_proportions.csv, deg_by_domain.csv with columns documented in the README.

Code

  • Standalone, commented R and Python scripts per analysis stage
  • Environment lock files: sessionInfo(), conda env export, or pip freeze
  • Delivery via private Git repository or agreed file transfer

Documentation

  • HTML/PDF QC report; README with reproduction steps; Methods draft with software versions
  • Post-delivery reviewer support: methods clarification and minor revisions within agreed scope (typically ≤20% of deliverables)

Technical decisions we make — and why

Spatially variable features: Squidpy Moran's I
Default for Visium and Xenium; alternative SVG methods on request for large cell-level datasets [3, 11].
Domain identification: BANKSY
Uses expression and neighborhood context together for tissue domains; graph-based spatial clustering methods on request [4, 13].
Deconvolution: cell2location when a matched scRNA reference exists
Models count data directly and handles batch covariates; SPOTlight for R-native workflows; reference-free deconvolution on request [6, 14, 15].
Differential expression: pseudobulk limma-voom with ≥3 replicates
Preferred over spot-level tests when replicate count permits; Wilcoxon spot-level only when pseudobulk is not feasible [1].
Batch correction: Harmony for same-platform cohorts
Experimental blocking and randomization preferred at study design; rigid coordinate alignment across slices scoped separately when 3D reconstruction is required [16].

Common questions

What is the minimum sequencing depth and number of spots for Visium spatial transcriptomics analysis?

For 10x Visium CytAssist v2, Pepkio recommends ≥25,000 read pairs per tissue-covered spot [2]. A 6.5 mm capture area contains 4,992 spots; effective spot count depends on tissue coverage. Visium HD minimum depth scales with preservation method, ≥275M read pairs for FFPE at 100% coverage on 6.5 mm [2]. At least three biological replicates per condition support pseudobulk differential expression; fewer samples can be analyzed with spot-level tests at reduced power.

Can you analyze low-quality FFPE sections or partially covered capture areas?

Yes, with caveats in the QC report. Partial coverage reduces usable spots; depth recommendations scale by capture-area fraction [2]. Degraded FFPE shows lower genes per spot and higher mitochondrial fractions; sub-threshold sections are flagged before domain calling.

Do you support 10x Visium, Visium HD, Xenium, CosMx, and custom analyses?

Yes. Visium v1/v2 and Visium HD use Space Ranger 3.1.3 from Illumina or MGI DNBSEQ-sequenced FASTQs when scoped at kickoff. Xenium outputs (Prime 5K and smaller panels) are imported with segmentation QC per Salas et al. [3]. CosMx and other platforms are supported on request with client-provided count matrices and coordinates. Custom work is scoped at kickoff.

How long does spatial transcriptomics analysis take at Pepkio?

A standard Visium cohort (4-12 sections, one tissue type, reference deconvolution) typically completes in 4-6 weeks from data receipt. Visium HD bin2cell, multi-slice integration, or Xenium segmentation review may take 6-10 weeks. Timelines are confirmed at kickoff.

How do you handle batch effects across patients, slides, or sequencing runs?

Experimental blocking and randomization are preferred at study design. For same-platform cohorts, Harmony corrects technical batches when shared spatial domains are present [16]. Donor, age, and preservation method stay in metadata and are included in pseudobulk models. Cross-platform or cross-slide 3D alignment is scoped separately when rigid coordinate registration is required.

Do I own the code, and in what format is it delivered?

Yes. You retain full ownership of all code, scripts, and results. Pepkio delivers commented R/Python scripts and environment lock files (sessionInfo(), conda, or pip). Objects use standard .h5ad and .rds formats readable in Scanpy or Seurat; R Markdown or Jupyter delivery is available on request.

Can I be involved during analysis?

Yes. Checkpoint reviews occur after QC, domain calling, and deconvolution mapping. You can review domain labels on H&E, adjust clustering resolution, and request contrasts within agreed scope. A dedicated scientific contact leads the project and incorporates your tissue-specific knowledge.

What does post-delivery reviewer support include?

Support covers clarification of computational methods, QC thresholds, and minor figure or table revisions within agreed scope (typically ≤20% of deliverables). Pepkio drafts Methods and Supplementary text for analyses we performed. Substantial new analyses requested by reviewers are scoped separately.

Is co-authorship required?

No. Pepkio operates as a fee-for-service provider unless co-authorship is explicitly discussed in advance. Standard practice is acknowledgment of bioinformatics support in the Acknowledgments section.

Do I need a scRNA-seq reference dataset for spatial deconvolution?

Not always. Reference-based deconvolution (cell2location, SPOTlight) requires a scRNA-seq atlas from comparable tissue that includes all major cell types present in the section [6]. Domain-level spatial statistics and SVG identification do not require a reference. Reference-free deconvolution is available when no suitable atlas exists; cell-type claims are then interpreted with appropriate caution.

Should I choose Visium spot-level or Visium HD cell-level analysis?

Visium v2 (55 µm spots, multiple cells per spot) suits tissue-domain discovery, niche architecture, and spatial gradients [1, 2]. Visium HD (2 µm bins, cell reconstruction via bin2cell on request) suits questions that need near-single-cell spatial resolution within histological context [17]. Platform choice is usually easier to settle before library prep; Pepkio can advise at kickoff based on your biological question.

Can you integrate my spatial data with an existing scRNA-seq atlas?

Yes, as a scoped milestone. Pepkio can train deconvolution signatures from your scRNA-seq reference and validate abundances against known markers [6, 14]. Joint Visium-Xenium integration on matched sections is available when both datasets exist.

Related services

  • Single-cell RNA-seq — Build or refine the scRNA-seq reference atlas used for spatial deconvolution and cell-type validation.
  • Bulk RNA-seq — Validate spatial pseudobulk signatures against independent cohort-level expression data.
  • Long-read RNA-seq — Isoform-level context for probe-based spatial platforms and splice-aware marker selection.
  • Multi-omics integration — Joint analysis of spatial transcriptomics with proteomics, CITE-seq, or matched imaging modalities.
  • Custom consulting — Platform selection, sequencing depth, and sectioning strategy before library prep.
References
  1. Moses L, Pachter L. Museum of spatial transcriptomics. Nature Methods. 2022;19(5):534–546. https://doi.org/10.1038/s41592-022-01409-2 (PMID: 35273392)
  2. 10x Genomics. Sequencing requirements for Visium CytAssist Spatial Gene Expression. 2024. https://www.10xgenomics.com/support/cytassist-spatial-gene-expression/documentation/steps/sequencing/sequencing-requirements-for-visium-cyt-assist-spatial-gene-expression-for-ffpe
  3. Salas SM, Kanemaru K, Tanaka H, et al. Optimizing Xenium In Situ data utility by quality assessment and best-practice analysis workflows. Nature Methods. 2025;22(4):706–716. https://doi.org/10.1038/s41592-025-02617-2 (PMID: 40082609)
  4. Long Y, Ang KS, Li Y, et al. Benchmarking clustering, alignment, and integration methods for spatial transcriptomics. Genome Biology. 2024;25(1):141. https://doi.org/10.1186/s13059-024-03361-0 (PMID: 39123269)
  5. Ståhl PL, Salmén F, Vickovic S, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78–82. https://doi.org/10.1126/science.aaf2403 (PMID: 27365449)
  6. Williams CG, Lee HJ, Bartoschek M, et al. An introduction to spatial transcriptomics for biomedical research. Genome Medicine. 2022;14(1):68. https://doi.org/10.1186/s13073-022-01075-1 (PMID: 35761361)
  7. Arora R, Cao C, Kumar M, et al. Spatial transcriptomics reveals distinct and conserved tumor core and edge architectures that predict survival and targeted therapy response. Nature Communications. 2023;14(1):4529. https://doi.org/10.1038/s41467-023-40271-4 (PMID: 37596273)
  8. Takano Y, Suzuki J, Nomura K, et al. Spatially resolved gene expression profiling of tumor microenvironment reveals key steps of lung adenocarcinoma development. Nature Communications. 2024;15(1):10637. https://doi.org/10.1038/s41467-024-54671-7 (PMID: 39639005)
  9. Wang X, Venet D, Lifrange F, et al. Spatial transcriptomics reveals substantial heterogeneity in triple-negative breast cancer with potential clinical implications. Nature Communications. 2024;15(1):8896. https://doi.org/10.1038/s41467-024-54145-w (PMID: 39592577)
  10. Grases D, Garcia ME, Navarro A, et al. A practical guide to spatial transcriptomics: lessons from over 1000 samples. Trends in Biotechnology. 2026;44(5):456–471. https://doi.org/10.1016/j.tibtech.2025.08.020 (PMID: 40975650)
  11. Palla G, Spitzer H, Klein M, et al. Squidpy: a scalable framework for spatial omics analysis. Nature Methods. 2022;19(2):171–178. https://doi.org/10.1038/s41592-021-01358-2 (PMID: 35102346)
  12. Ni Z, Prasad A, Chen S, et al. SpotClean adjusts for spot swapping in spatial transcriptomics data. Nature Communications. 2022;13(1):2971. https://doi.org/10.1038/s41467-022-30587-y (PMID: 35624112)
  13. Singhal V, Chou N, Lee J, et al. BANKSY unifies cell typing and tissue domain segmentation for scalable spatial omics data analysis. Nature Genetics. 2024;56(3):431–441. https://doi.org/10.1038/s41588-024-01664-3 (PMID: 38413725)
  14. Kleshchevnikov V, Shmatko A, Dann E, et al. Cell2location maps fine-grained cell types in spatial transcriptomics. Nature Biotechnology. 2022;40(5):661–671. https://doi.org/10.1038/s41587-021-01139-4 (PMID: 35027729)
  15. Elosua-Bayes M, Nieto P, Mereu E, et al. SPOTlight: seeded NMF regression to deconvolute spatial transcriptomics spots with single-cell transcriptomes. Nucleic Acids Research. 2021;49(9):e50. https://doi.org/10.1093/nar/gkab043 (PMID: 33544846)
  16. Korsunsky I, Millard N, Fan J, et al. Fast, sensitive and accurate integration of single-cell data with Harmony. Nature Methods. 2019;16(12):1289–1296. https://doi.org/10.1038/s41592-019-0619-0 (PMID: 31740819)
  17. Polanski K, Young MD, Miao Z, et al. Bin2cell reconstructs cells from high-resolution Visium HD data. Bioinformatics. 2024;40(9):btae546. https://doi.org/10.1093/bioinformatics/btae546 (PMID: 39250728)

Let's Talk About Your Science

Tell us:

  • • Your biological question
  • • Data type and size
  • • Timeline constraints

We'll tell you:

  • • What's feasible
  • • How long it will take
  • • Exactly what it will cost
Discuss Your Project

Contact us to start with a free consultation. Need everyday bench calculators? Try our free lab tools.