https://orcid.org/0000-0003-3797-2148
Figures
Abstract
Spatial transcriptomics (ST) enables genome-wide gene expression profiling while retaining spatial context within tissue sections. Since the foundational work by Ståhl et al. in 2016, the field has expanded rapidly, with diverse platforms now spanning sequencing-based (e.g., Visium, Visium HD, Slide-seq, Stereo-seq, and Seq-Scope) and imaging-based (e.g., MERFISH, Xenium, and CosMx SMI) approaches. The breadth of platforms, data structures, and computational tools, however, can be daunting for newcomers. Here, we present ten quick tips spanning the entire ST research workflow: whether ST suits a given biological question, how to select a platform aligned with study objectives, how to understand and process ST data, and which software tools to employ for analysis and visualization. We further discuss interpreting spatial patterns in biological context, integrating complementary modalities such as single-cell RNA sequencing and spatial proteomics, and leveraging public datasets and sharing results. Finally, we highlight current limitations of ST, particularly the challenge of reconstructing three-dimensional tissue architecture from serial tissue sections. This review provides biologists, bioinformaticians, and clinician-scientists with a concise, platform-neutral roadmap for incorporating ST into research, from experimental design to biological discovery.
Author summary
Spatial transcriptomics (ST) lets researchers measure gene expression while preserving the tissue location of each measurement. Where a cell sits, and which cells surround it, shape how it behaves—information that is lost when tissue is dissociated. Despite the growing popularity of ST, there is currently no concise, practical guide for researchers new to the field, for whom the diversity of platforms, data formats, and computational tools can be overwhelming. Here, we distill essential knowledge into ten quick tips covering the entire ST research workflow—from deciding whether ST is the right approach for a given biological question, through selecting an appropriate experimental platform, to analyzing and interpreting the resulting data. We also discuss integrating ST with other omics data and highlight current limitations that users should be aware of. Our goal is to help researchers across disciplines—including oncology, neuroscience, and developmental biology—design rigorous ST experiments, avoid common pitfalls, and generate reproducible, biologically meaningful insights.
Citation: Kurogi N, Shimbara K, Koreeda T, Tsuyuzaki K (2026) Ten quick tips for spatial transcriptomics analysis. PLoS Comput Biol 22(9): e1014757. https://doi.org/10.1371/journal.pcbi.1014757
Editor: Francis Ouellette, Montreal, CANADA
Published: September 15, 2026
Copyright: © 2026 Kurogi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: This article does not report new data. All information discussed is available in the cited references.
Funding: This work was supported by JSPS KAKENHI (grant numbers 23K11312, 24K02011, 24K21269, and 23K23933 to KT) and JST A-STEP (grant number JPMJTR25UJ to KT). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Spatial transcriptomics (ST) comprises a family of technologies that measure gene expression in tissue sections while preserving spatial coordinates. In contrast to dissociation-based approaches such as single-cell RNA sequencing (scRNA-seq), ST links molecular measurements to tissue architecture, enabling researchers to ask where cell types, gene-expression programs, and disease-associated niches occur within their histological context [1].
ST is rapidly establishing itself as a tool for relating tissue structure to function across fields including oncology [2], neuroscience [3], and developmental biology [4]. As in other omics assays, the questions an ST study can answer are constrained well before analysis begins—but these constraints bind unusually tightly here: tissue preservation and handling, the spatial resolution and transcriptome coverage of the chosen platform, the available histological context, and the number of biological replicates jointly determine which spatial hypotheses are answerable. What is specific to ST is the statistical structure of the resulting data: neighboring measurements are spatially autocorrelated rather than independent, and the many spots in a section do not substitute for biological replication. Planning therefore means reasoning explicitly about these spatial and statistical properties from the outset, not only about cost or throughput.
In this article, we present ten quick tips as practical guidance for researchers employing ST, covering the entire research workflow from platform selection and data understanding to biological interpretation, multi-omics integration, and recognition of technical limitations.
Tip 1: Confirm that spatial information is essential before committing to ST
Although ST is an innovative alternative to conventional experimental approaches such as bulk RNA-seq, it is not necessarily the optimal method for every research question. Before initiating a study, ask: is spatial information truly indispensable for addressing the biological question at hand? ST is most valuable when the spatial arrangement of gene expression is itself part of the question—for example, consider whether the overall question requires information on the organization of cell types, tissue architecture, local microenvironments and niches, signaling gradients, or cell–cell interactions interpreted in their tissue setting [2,4]. ST experiments are relatively costly per assay, and the resulting data are complex, requiring specialized expertise and time to analyze [5]. If the objective can be achieved with established methods such as scRNA-seq, immunohistochemistry (IHC), or in situ hybridization (ISH), consider whether ST is cost-effective [6].
This is especially relevant for large-scale screening, where ST throughput and cost remain limiting. ST is best suited to investigations seeking deep insight from a limited number of precious specimens—for example, dissecting microenvironment architecture in specific disease models [7]—whereas evaluating a predefined biomarker across hundreds of clinical specimens is typically far more efficient with IHC. Cost and throughput are real constraints, but they should be weighed as one consideration among several rather than treated as the primary criterion.
Finally, define the comparison and replication strategy before generating data. Decide which groups will be compared (e.g., disease versus control, or across anatomical regions) and build in biological replication at the inferential level—patients, animals, or samples—not the many spots within a single section (Tip 4). Decide also how much tissue to profile: small cores or tissue microarrays raise throughput but capture limited spatial context, whereas large sections preserve architecture at the cost of fewer samples and higher per-sample cost. In case–control designs, balance conditions across slides and batches, or randomize their placement, so that biological contrasts are not confounded with technical ones (Tip 5). Pilot experiments and sample-size reasoning here help establish whether an ST study can realistically support its conclusions.
Tip 2: Choose an ST platform that fits your biological question and resources
ST platforms fall into two broad families that differ in their underlying chemistry (Table 1; Fig 1). Imaging-based (in situ hybridization) approaches detect individual transcripts in place using sequential rounds of fluorescent probe hybridization and imaging; representative methods include single-molecule FISH (smFISH), seqFISH, and MERFISH [8], with commercial implementations such as Vizgen MERSCOPE, NanoString CosMx SMI [9], and 10x Genomics Xenium [10]. Sequencing-based (spatial barcoding or capture) approaches capture transcripts on spatially barcoded arrays or beads and read them out by sequencing; representative methods include Visium and Visium HD [1,11], Slide-seq [12], Stereo-seq [13], Seq-Scope [14], the bead-array Curio Trekker, and the region-of-interest–based GeoMx Digital Spatial Profiler (DSP).
Imaging-based platforms and sequencing-based platforms are plotted by spatial resolution (x-axis, µm) versus the number of genes detected per spot or cell (y-axis). The same two-family classification is used in Table 1. Representative imaging-based methods include smFISH, seqFISH, MERFISH/MERSCOPE, CosMx SMI, and Xenium; representative sequencing-based methods include Visium, Visium HD, Slide-seq, Stereo-seq, Seq-Scope, Curio Trekker, and GeoMx DSP. Shading distinguishes subcellular resolution (<10 µm) from resolution ≥10 µm.
The two families trade resolution against breadth in opposite directions. Imaging-based methods reach subcellular resolution and high sensitivity but typically profile a predefined panel of tens to a few thousand genes, suiting hypothesis-testing on targeted gene sets. Sequencing-based methods measure the transcriptome genome-wide, suiting exploratory discovery of uncharacterized spatial programs, but historically at coarser, multi-cell resolution. That gap is narrowing: Visium HD tiles ~2-µm capture areas toward near-single-cell resolution [11], Stereo-seq uses DNA-nanoball arrays at submicron spacing with adjustable bin size [13], and Seq-Scope exploits Illumina flow-cell clusters for submicron barcoding [14]. Resolution alone, however, does not determine the effective single-cell information recovered, which also depends on capture efficiency, transcript density, and segmentation quality (Tip 5).
Sample type often constrains platform choice as much as the biological question does, and these requirements are largely imposed by the platform rather than freely chosen. Academic platforms at high resolution (e.g., Stereo-seq, Seq-Scope) generally require fresh-frozen tissue, and thus prospective collection and snap-freezing. Probe- or capture-based targeted platforms (e.g., Visium FFPE, Xenium, CosMx) instead accept formalin-fixed paraffin-embedded (FFPE) material—making them applicable to archival and retrospective clinical cohorts—but, because FFPE RNA is fragmented and chemically modified, they depend on probe/capture chemistries rather than poly-A whole-transcriptome capture [6]. Identifying a platform’s required fixation and preservation conditions early, and aligning them with available samples, helps establish a feasible workflow and improves the reliability of the results [6]. Targeted platforms also differ in whether custom gene panels can be designed for a given tissue or organism, and some imaging-based assays have been explored for spatial detection of selected sequence variants; because this is platform- and assay-design-dependent, confirm that the specific variants, tissue, and probe design are supported before building variant detection into the design [15].
Finally, weigh commercial against noncommercial platforms. Commercial systems (e.g., Visium, Xenium, and CosMx) offer standardized reagents and support but carry substantial instrument and per-sample costs, whereas academic methods (e.g., Slide-seq, Stereo-seq, and Seq-Scope) can lower reagent costs or push resolution while demanding more in-house expertise [6]. The most expensive or highest-resolution platform is not automatically the best choice; recent independent benchmarks comparing commercial and academic ST platforms can inform this trade-off [6,16].
Tip 3: Familiarize yourself with the structure of ST data before analysis
ST datasets comprise two distinct modalities: (i) tissue images (e.g., H&E or fluorescence) and (ii) spatially resolved gene-expression measurements linked to those images. Regardless of platform, most share a common set of components: an expression matrix of features (genes) by observations (spots, cells, or bins); spatial coordinates for each observation; feature/gene metadata; observation-level metadata; and one or more registered tissue images where available. The first analytical step is to generate the expression matrix of genes × spatial coordinates (spots or cells) from raw data (Fig 2): in sequencing-based methods by mapping reads that carry spatial barcodes, and in imaging-based methods by decoding multi-round fluorescence images.
The project root contains the gene expression × spot matrix (filtered_feature_bc_matrix.h5) and a Spatial/subdirectory holding spatial coordinate files (Tissue_positions.csv), scaling parameters (Scalefactors_json.json), tissue images at multiple resolutions (Tissue_hires_image.png, Tissue_lowres_image.png), and sample metadata (Metadata.csv). Other platforms emit different files and conventions. Icons used in this figure were adapted from IconaMoon by Dariush Habibpour, obtained via Iconify, and are licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0). The icons were modified in color and appearance.
The precise file organization is platform-specific. Taking 10x Visium as one common example, output is organized into a feature–barcode count matrix (e.g., filtered_feature_bc_matrix.h5) alongside a spatial directory holding tissue-position coordinates, scale factors, and tissue images at multiple resolutions (Fig 2). Other platforms—Slide-seq, Stereo-seq, Seq-Scope, or imaging-based assays—produce different files, so the Visium layout in Fig 2 is illustrative rather than a universal standard.
Where multiple cells occupy a single spot, image-based segmentation can estimate cell boundaries and reconstruct cell-level profiles [17]. When a reference scRNA-seq dataset is available (Tip 8), cell-type deconvolution can estimate the proportions of cell types within each spot [18].
For data management and reproducibility, standardized multi-layer formats—such as AnnData/H5AD (Python) [19] and Seurat objects (R) [20]—are recommended; these store the expression matrix, spatial coordinates, images, cell/gene metadata, and downstream results in a single file, facilitating integrated analysis and sharing.
Tip 4: Follow a structured workflow to analyze ST data systematically
Once the expression matrix is constructed, analysis proceeds through several stages that extend conventional scRNA-seq workflows with spatial information [20] (Fig 3). No single toolkit is required; those named below are examples, and Table 2 compares options so readers can choose by task rather than ecosystem.
Stages: (1) experimental design and metadata curation (Tips 1–2); (2) data acquisition by sequencing-based (capture) or imaging-based methods (Tip 2); (3) import, quality control, and normalization (Tips 3–5; e.g., Scanpy, Seurat, sctransform); (4) dimensionality reduction and clustering (Tip 4; e.g., Scanpy, Seurat); (5) spatial analysis—spatial-domain detection, spatially variable genes, deconvolution, and cell–cell communication inference (Tips 4–8; e.g., Squidpy, Giotto, SpatialDE, cell2location, NicheNet); and (6) biological interpretation and integration (Tips 7–8). Tool names are examples, not recommendations; see Table 2. Numbers in brackets indicate the most relevant tip(s).
Preprocessing
Perform quality control (QC) to remove low-quality spots or cells (e.g., extremely low gene counts, abnormally high mitochondrial fractions), and normalize for differences in library size (total counts per spot). Because ST counts can be highly overdispersed, variance-stabilizing approaches such as sctransform (in Seurat) are often employed [21]. Because imaging-based and sequencing-based platforms generate counts differently, normalization and variance modeling should reflect platform chemistry, capture efficiency, panel design, and tissue type rather than be copied from scRNA-seq workflows.
Core analysis
As in scRNA-seq, apply dimensionality reduction (e.g., principal component analysis) and clustering to partition the tissue into regions with similar expression profiles, characterizing tissue architecture and cellular distributions [20,22]. These steps ignore spatial coordinates and are therefore a starting point, not a spatial analysis.
Spatially informed analysis
This is where ST’s strengths are fully leveraged. Representative analyses include identifying spatially variable genes (SVGs) with layered, gradient, or patch-like distributions [23]; inferring cell–cell communication from the spatial co-expression of ligands and receptors [24]; and detecting spatial domains (niches) via approaches such as neighborhood enrichment [25]. Others include spatial autocorrelation (e.g., Moran’s I), cell-type deconvolution of multi-cell spots against a reference (Tip 8), and segmentation-aware analysis tying expression to cell boundaries and histology (Tip 5). These reveal functional tissue units and aberrant intercellular signaling in disease. Because domain detection, SVG calling, and communication inference are each served by several methods with different assumptions (Table 2), results should be checked for robustness to method and parameter choice.
Respect the statistical structure of the data
Spots or cells within a section are spatially autocorrelated and are not independent biological replicates; treating thousands of spots as the sample size inflates significance and constitutes pseudoreplication [26]. For group comparisons and differential expression, replication should be counted at the biological level (patient, animal, or sample), and spatial dependence modeled rather than ignored. Where conditions are confounded with slides or batches, observed differences can reflect technical rather than biological effects (Tip 5).
Tip 5: Use dedicated tools for visualizing and segmenting spatial images
High-resolution tissue images in ST are invaluable, and inspecting histological context alongside gene-expression maps is indispensable. Platform vendors provide GUI applications, such as Loupe Browser (10x Genomics), that allow interactive visualization of gene expression overlaid on images without programming, enabling rapid assessment of data quality and gene localization.
Visual inspection is also the primary defense against tissue and image artifacts that distort downstream results. Watch for tissue folds, tears, detachment, and mounting or bubble artifacts. Watch also for regions of low RNA quality caused by degradation, suboptimal fixation, or—in fresh-frozen workflows—freezing and OCT-embedding damage. Sequencing-based assays additionally depend on permeabilization, which must be optimized for each tissue because over- or under-permeabilization biases capture [6,27]. Affected regions should be flagged or excluded during QC rather than interpreted as biology, and their distribution should be checked against slides and batches so that artifacts are not mistaken for condition effects (Tips 1 and 4).
For more advanced image analysis—such as machine-learning-based cell detection and the quantitative assessment of pathological features—open-source tools are powerful options. QuPath [28], a pathology image-analysis platform, and napari [29], a Python-based multidimensional image viewer, can support integration with ST data. Combining these environments with Cellpose, a general-purpose deep-learning-based segmentation tool, can help shift sequencing-based ST analysis from spot-level resolution toward single-cell-level spatial analysis by enabling the reconstruction of high-resolution cellular maps [17].
Segmentation quality depends on both the experiment and the analysis. Experimentally, nuclear and membrane staining determines how clearly cell boundaries appear in the image; computationally, deep-learning tools such as Cellpose [17] use those boundaries to assign transcripts to individual cells. When either step fails, cells are merged or split and transcripts are assigned to the wrong cell, and these errors carry through into per-cell expression estimates. Segmentation should therefore be inspected and validated rather than assumed to be correct, especially in dense or irregularly shaped tissue.
Tip 6: Choose specialized software packages for ST gene-expression analysis
Effective analysis of ST gene-expression data typically relies on software packages designed for spatial information (Table 2). Such toolkits provide spatially aware data structures, specialized analytical algorithms, and visualization utilities, thereby substantially improving analytical efficiency and reproducibility. The right choice depends on the analysis at hand and the language a group already works in; Table 2 compares representative tools by supported analysis, strengths, and limitations.
In the R ecosystem, the scRNA-seq framework Seurat has integrated spatial analysis capabilities and is widely adopted [20], while Giotto [30] and STUtility [31] offer rich, ST-focused functionality. In Python, Scanpy [22] together with its extension Squidpy [25] is broadly used; Squidpy implements spatial statistics including construction of spatial graphs, neighborhood-enrichment analyses, and ligand–receptor interaction inference. Beyond these core packages, BayesSpace [32] incorporates spatial priors in a Bayesian framework to upsample spot-level data toward subspot resolution. As pipelines grow in complexity, encoding the analysis with workflow managers such as Nextflow or Snakemake is critical for ensuring reproducibility and facilitating collaboration.
Tip 7: Ground your interpretation of spatial patterns in biological knowledge
The most important—and creative—stage of ST analysis is translating observed data patterns into coherent biological narratives. Heat maps and clustering gain scientific value only when anchored in biological meaning; interpreting the spatial localization of specific genes or the distribution of newly identified cell populations requires deep domain knowledge and systematic cross-referencing with the literature and relevant databases.
Several landmark studies exemplify this process. In Alzheimer’s disease, Chen et al. mapped gene expression around amyloid plaques and identified inflammatory programs in proximal glial cells [7]. In oncology, Ravi et al. integrated spatial transcriptomics and proteomics to reveal spatially segregated glioblastoma microenvironments with distinct metabolic and immune states [2]. In developmental biology, Asp et al. produced a spatiotemporal atlas of the human fetal heart, detailing cell-type distributions and gene-expression dynamics underlying structures such as the conduction system [4]. In each case, knowing where each measurement was taken—information that tissue dissociation discards—was essential to the insight. The shared lesson is to anchor interpretation in the tissue: align clusters and spatial domains to histology and anatomical landmarks and confirm domain identity with known marker genes before naming it.
Spatial proximity and inferred interaction also invite over-interpretation. Co-localization of two cell types, or co-expression of a ligand and its receptor, is consistent with interaction but does not by itself demonstrate that the cells communicate or that one state causes another; inferred cell–cell communication is best treated as hypothesis generation to be tested experimentally [24]. Therefore, hypotheses emerging from ST should, where possible, be validated against complementary data. Large language models (LLMs) have recently been explored for assisting cell-type annotation in single-cell omics [39]; while early results show moderate accuracy, such approaches require caution and expert oversight.
Tip 8: Integrate ST with complementary omics data to deepen your findings
ST becomes more informative when integrated with complementary data. Integration with scRNA-seq is now routine: a reference atlas of cell types is mapped onto a section to estimate which cell types are present and in what proportions, giving cell-type-level interpretation of spot-based measurements rather than true single-cell resolution. Representative methods include cell2location [18], a Bayesian model of cell-type composition within each spot, and Tangram [35], which aligns single-cell profiles to spatial data with deep learning. Such integration is only as good as the reference: one that omits a cell type or derives from a different condition or technology can bias the mapping, so reference choice should match the tissue and question.
Beyond transcriptomes, ST can be combined with other spatial modalities for a multi-layered view of tissue. Spatial proteomics platforms such as CODEX or imaging mass cytometry (IMC) add protein-level readouts in the same frame, testing whether mRNA abundance tracks protein function [40,41]; DBiT-seq co-maps mRNA and protein from one section [42]. Emerging spatial epigenomic and metabolomic assays extend this to chromatin state and small-molecule distributions [43,44], complementary layers that transcriptomes alone cannot provide.
Integration is not automatically informative. Modalities differ in resolution, are often measured on different sections or platforms, and may require registration; resolution mismatch, batch effects, and alignment uncertainty can introduce artifacts that masquerade as biology. Before integrating, decide what each modality adds, verify registration, and confirm the combined analysis improves interpretation rather than merely adding complexity.
Tip 9: Leverage public ST datasets and practice open science
ST is a rapidly evolving, interdisciplinary field in which cross-domain learning is highly beneficial. In neuroscience, the Allen Brain Map systematically relates brain structure to gene expression, providing gold-standard references and benchmarks for spatial analyses [45]; in developmental biology, a Stereo-seq-based atlas of Drosophila has produced three-dimensional spatiotemporal transcriptomic maps of embryonic and larval development, inspiring new analytical methods and concepts [46].
Many datasets are accessible through public repositories such as SpatialDB [47] and STOmicsDB [48] (Table 3), enabling preliminary hypothesis testing and method learning. Practicing open science—sharing code, metadata, and raw/processed data alongside manuscripts—enhances transparency and accelerates community-wide progress. For ST specifically, this means depositing raw and processed data (expression matrix, spatial coordinates, and tissue images) in standard, interoperable formats such as AnnData/H5AD; recording metadata on platform, tissue handling, and experimental design; sharing analysis code as versioned scripts or notebooks that state software versions and key parameters; and releasing data and code under permissive licenses that meet the target journal’s data-availability requirements so others can reanalyze and compare results.
Tip 10: Acknowledge the current technical limitations of ST
Like any technology, ST has current limitations that must be acknowledged to ensure rigorous interpretation.
Several limitations are technical and experimental. Sensitivity and detection efficiency are typically lower than bulk RNA-seq, so lowly expressed genes may be missed; spatial resolution varies by platform, and at multi-cell resolution each spot mixes signals from several cells, which deconvolution only partially resolves. Segmentation errors, tissue-processing artifacts (Tip 5), and batch effects between slides or runs further limit comparability, and cross-platform comparison remains difficult because each platform carries distinct biases [6]. Other limitations are statistical and computational: spatial autocorrelation and pseudoreplication complicate inference (Tip 4), and many spatial methods carry their own modeling assumptions and parameter sensitivity. Finally, ST is largely observational—co-localization and inferred interactions support hypotheses but rarely establish causation (Tip 7), so spatial associations should be framed as hypotheses unless supported by perturbation, orthogonal validation, or independent replication.
A further, much-discussed limitation is dimensionality: most ST methods rely on thin tissue sections and thus yield inherently two-dimensional measurements. Reconstructing true three-dimensional tissue architecture from serial sections introduces substantial technical challenges, and for experiments spanning many sections no universally accepted, high-precision method exists for aligning independently measured sections. In practice, researchers often combine multiple image-registration tools (e.g., Fiji/ImageJ, ANTs, and Elastix) and iteratively tune parameters by hand. Although nonrigid registration, optimal transport, and deep learning approaches have been proposed, heterogeneity in I/O formats, parameterization, and workflow integration has impeded robust, automated standardization, so many studies still analyze sections individually and qualitatively compare expression maps across sections.
Current algorithms such as PASTE [38] formulate alignment as an optimal-transport problem that balances transcriptional similarity with physical distance between adjacent sections. A recent benchmark of 12 multi-slice integration methods across 19 datasets found that no single method dominates universally; performance depends strongly on dataset characteristics, technology, and downstream goals [49]. These findings reinforce the need for context-aware method choice, transparent reporting of integration settings, and sensitivity analyses when reconstructing 3D tissue architecture.
Conclusions
Spatial transcriptomics now provides a rapidly advancing framework for mapping gene expression in its native tissue context. This article consolidates practical guidance to help researchers formulate spatially explicit hypotheses, align platform selection and experimental design with those hypotheses, implement reproducible, spatially aware analytical workflows, and integrate orthogonal modalities to validate and extend findings. Because ST is observational, the associations it reveals should be treated as hypotheses. Adherence to these principles can help turn spatial maps into testable models of tissue organization and its perturbation in disease, validated with orthogonal and experimental data.
References
- 1. Ståhl PL, Salmén F, Vickovic S, Lundmark A, Navarro JF, Magnusson J, et al. Visualization and analysis of gene expression in tissue sections by spatial transcriptomics. Science. 2016;353(6294):78–82. pmid:27365449
- 2. Ravi VM, Will P, Kueckelhaus J, Sun N, Joseph K, Salié H, et al. Spatially resolved multi-omics deciphers bidirectional tumor-host interdependence in glioblastoma. Cancer Cell. 2022;40(6):639-655.e13. pmid:35700707
- 3. Zhang M, Eichhorn SW, Zingg B, Yao Z, Cotter K, Zeng H, et al. Spatially resolved cell atlas of the mouse primary motor cortex by MERFISH. Nature. 2021;598(7879):137–43. pmid:34616063
- 4. Asp M, Giacomello S, Larsson L, Wu C, Fürth D, Qian X, et al. A spatiotemporal organ-wide gene expression and cell atlas of the developing human heart. Cell. 2019;179(7):1647-1660.e19. pmid:31835037
- 5. Marx V. Method of the Year: spatially resolved transcriptomics. Nat Methods. 2021;18(1):9–14. pmid:33408395
- 6. Lim HJ, Wang Y, Buzdin A, Li X. A practical guide for choosing an optimal spatial transcriptomics technology from seven major commercially available options. BMC Genomics. 2025;26(1):47. pmid:39833687
- 7. Chen W-T, Lu A, Craessaerts K, Pavie B, Sala Frigerio C, Corthout N, et al. Spatial transcriptomics and in situ sequencing to study Alzheimer’s disease. Cell. 2020;182(4):976-991.e19. pmid:32702314
- 8. Chen KH, Boettiger AN, Moffitt JR, Wang S, Zhuang X. RNA imaging. Spatially resolved, highly multiplexed RNA profiling in single cells. Science. 2015;348(6233):aaa6090. pmid:25858977
- 9. He S, Bhatt R, Brown C, Brown EA, Buhr DL, Chantranuvatana K, et al. High-plex imaging of RNA and proteins at subcellular resolution in fixed tissue by spatial molecular imaging. Nat Biotechnol. 2022;40(12):1794–806. pmid:36203011
- 10. Janesick A, Shelansky R, Gottscho AD, Wagner F, Williams SR, Rouault M, et al. High resolution mapping of the tumor microenvironment using integrated single-cell, spatial and in situ analysis. Nat Commun. 2023;14(1):8353. pmid:38114474
- 11. Oliveira MF de, Romero JP, Chung M, Williams SR, Gottscho AD, Gupta A, et al. High-definition spatial transcriptomic profiling of immune cell populations in colorectal cancer. Nat Genet. 2025;57(6):1512–23. pmid:40473992
- 12. Rodriques SG, Stickels RR, Goeva A, Martin CA, Murray E, Vanderburg CR, et al. Slide-seq: a scalable technology for measuring genome-wide expression at high spatial resolution. Science. 2019;363(6434):1463–7. pmid:30923225
- 13. Chen A, Liao S, Cheng M, Ma K, Wu L, Lai Y, et al. Spatiotemporal transcriptomic atlas of mouse organogenesis using DNA nanoball-patterned arrays. Cell. 2022;185(10):1777-1792.e21. pmid:35512705
- 14. Cho C-S, Xi J, Si Y, Park S-R, Hsu J-E, Kim M, et al. Microscopic examination of spatial transcriptome using Seq-Scope. Cell. 2021;184(13):3559-3572.e22. pmid:34115981
- 15. Zhuang X. Spatially resolved single-cell genomics and transcriptomics by imaging. Nat Methods. 2021;18(1):18–22. pmid:33408406
- 16. Ren P, Zhang R, Wang Y, Zhang P, Luo C, Wang S, et al. Systematic benchmarking of high-throughput subcellular spatial transcriptomics platforms across human tumors. Nat Commun. 2025;16(1):9232. pmid:41107232
- 17. Stringer C, Wang T, Michaelos M, Pachitariu M. Cellpose: a generalist algorithm for cellular segmentation. Nat Methods. 2021;18(1):100–6. pmid:33318659
- 18. Kleshchevnikov V, Shmatko A, Dann E, Aivazidis A, King HW, Li T, et al. Cell2location maps fine-grained cell types in spatial transcriptomics. Nat Biotechnol. 2022;40(5):661–71. pmid:35027729
- 19. Virshup I, Rybakov S, Theis FJ, Angerer P, Wolf FA. Anndata: access and store annotated data matrices. JOSS. 2024;9(101):4371.
- 20. Hao Y, Hao S, Andersen-Nissen E, Mauck WM 3rd, Zheng S, Butler A, et al. Integrated analysis of multimodal single-cell data. Cell. 2021;184(13):3573-3587.e29. pmid:34062119
- 21. Hafemeister C, Satija R. Normalization and variance stabilization of single-cell RNA-seq data using regularized negative binomial regression. Genome Biol. 2019;20(1):296. pmid:31870423
- 22. Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data analysis. Genome Biol. 2018;19(1):15. pmid:29409532
- 23. Svensson V, Teichmann SA, Stegle O. SpatialDE: identification of spatially variable genes. Nat Methods. 2018;15(5):343–6. pmid:29553579
- 24. Browaeys R, Saelens W, Saeys Y. NicheNet: modeling intercellular communication by linking ligands to target genes. Nat Methods. 2020;17(2):159–62. pmid:31819264
- 25. Palla G, Spitzer H, Klein M, Fischer D, Schaar AC, Kuemmerle LB, et al. Squidpy: a scalable framework for spatial omics analysis. Nat Methods. 2022;19(2):171–8. pmid:35102346
- 26. Hurlbert SH. Pseudoreplication and the design of ecological field experiments. Ecol Monogr. 1984;54(2):187–211.
- 27. Fang Y, Peng Z, Wang Y, Yuan X, Gao K, Fan R, et al. Improvements and challenges of tissue preparation for spatial transcriptome analysis of skull base tumors. Heliyon. 2023;9(3):e14133. pmid:36938455
- 28. Bankhead P, Loughrey MB, Fernández JA, Dombrowski Y, McArt DG, Dunne PD, et al. QuPath: open source software for digital pathology image analysis. Sci Rep. 2017;7(1):16878. pmid:29203879
- 29.
napari contributors. Napari: a multi-dimensional image viewer for Python; 2019. https://doi.org/10.5281/zenodo.3555620
- 30. Dries R, Zhu Q, Dong R, Eng C-HL, Li H, Liu K, et al. Giotto: a toolbox for integrative analysis and visualization of spatial expression data. Genome Biol. 2021;22(1):78. pmid:33685491
- 31. Bergenstråhle J, Larsson L, Lundeberg J. Seamless integration of image and molecular analysis for spatial transcriptomics workflows. BMC Genomics. 2020;21(1):482. pmid:32664861
- 32. Zhao E, Stone MR, Ren X, Guenthoer J, Smythe KS, Pulliam T, et al. Spatial transcriptomics at subspot resolution with BayesSpace. Nat Biotechnol. 2021;39(11):1375–84. pmid:34083791
- 33. Hu J, Li X, Coleman K, Schroeder A, Ma N, Irwin DJ, et al. SpaGCN: integrating gene expression, spatial location and histology to identify spatial domains and spatially variable genes by graph convolutional network. Nat Methods. 2021;18(11):1342–51. pmid:34711970
- 34. Cable DM, Murray E, Zou LS, Goeva A, Macosko EZ, Chen F, et al. Robust decomposition of cell type mixtures in spatial transcriptomics. Nat Biotechnol. 2022;40(4):517–26. pmid:33603203
- 35. Biancalani T, Scalia G, Buffoni L, Avasthi R, Lu Z, Sanger A, et al. Deep learning and alignment of spatially resolved single-cell transcriptomes with Tangram. Nat Methods. 2021;18(11):1352–62. pmid:34711971
- 36. Lopez R, Li B, Keren-Shaul H, Boyeau P, Kedmi M, Pilzer D, et al. DestVI identifies continuums of cell types in spatial transcriptomics data. Nat Biotechnol. 2022;40(9):1360–9. pmid:35449415
- 37. Dimitrov D, Türei D, Garrido-Rodriguez M, Burmedi PL, Nagai JS, Boys C, et al. Comparison of methods and resources for cell-cell communication inference from single-cell RNA-Seq data. Nat Commun. 2022;13(1):3224. pmid:35680885
- 38. Zeira R, Land M, Strzalkowski A, Raphael BJ. Alignment and integration of spatial transcriptomics data. Nat Methods. 2022;19(5):567–75. pmid:35577957
- 39. Hou W, Ji Z. Assessing GPT-4 for cell type annotation in single-cell RNA-seq analysis. Nat Methods. 2024;21(8):1462–5. pmid:38528186
- 40. Goltsev Y, Samusik N, Kennedy-Darling J, Bhate S, Hale M, Vazquez G, et al. Deep profiling of mouse splenic architecture with CODEX multiplexed imaging. Cell. 2018;174(4):968-981.e15. pmid:30078711
- 41. Giesen C, Wang HAO, Schapiro D, Zivanovic N, Jacobs A, Hattendorf B, et al. Highly multiplexed imaging of tumor tissues with subcellular resolution by mass cytometry. Nat Methods. 2014;11(4):417–22. pmid:24584193
- 42. Liu Y, Yang M, Deng Y, Su G, Enninful A, Guo CC, et al. High-spatial-resolution multi-omics sequencing via deterministic barcoding in tissue. Cell. 2020;183(6):1665-1681.e18. pmid:33188776
- 43. Deng Y, Bartosovic M, Ma S, Zhang D, Kukanja P, Xiao Y, et al. Spatial profiling of chromatin accessibility in mouse and human tissues. Nature. 2022;609(7926):375–83. pmid:35978191
- 44. Alexandrov T. Spatial Metabolomics and imaging mass spectrometry in the age of artificial intelligence. Annu Rev Biomed Data Sci. 2020;3:61–87. pmid:34056560
- 45. Yao Z, van Velthoven CTJ, Kunst M, Zhang M, McMillen D, Lee C, et al. A high-resolution transcriptomic and spatial atlas of cell types in the whole mouse brain. Nature. 2023;624(7991):317–32. pmid:38092916
- 46. Wang M, Hu Q, Lv T, Wang Y, Lan Q, Xiang R, et al. High-resolution 3D spatiotemporal transcriptomic maps of developing Drosophila embryos and larvae. Dev Cell. 2022;57(10):1271-1283.e4. pmid:35512700
- 47. Fan Z, Chen R, Chen X. SpatialDB: a database for spatially resolved transcriptomes. Nucleic Acids Res. 2020;48(D1):D233–7. pmid:31713629
- 48. Xu Z, Wang W, Yang T, Li L, Ma X, Chen J, et al. STOmicsDB: a comprehensive database for spatial transcriptomics data sharing, analysis and visualization. Nucleic Acid Res. 2024;52: D1053–D1061.
- 49. Dong K, Gao Y, Zou Q, Cui Y, Han C, Lin S, et al. Benchmarking multi-slice integration and downstream applications in spatial transcriptomics data analysis. Genome Biol. 2025;26(1):318. pmid:41024097