Spatial sequencing is a family of technologies that measure which genes are active in a tissue while preserving the physical location of every measurement. Traditional methods of studying gene activity require grinding tissue into a slurry, which reveals what genes cells are using but destroys any information about where those cells sat relative to one another. Spatial sequencing keeps the tissue intact, producing maps that show not just which genes are switched on but exactly where in the tissue that activity occurs. The field has grown rapidly in the past few years, and understanding the different approaches, their trade-offs, and what they have already revealed gives a clearer picture of where biology and medicine are headed.
Why Location Matters
A tumor is not a uniform mass. Its core, its invasive edges, and the immune cells scattered around it all behave differently. The same is true for a developing brain, a damaged kidney, or an inflamed gut. When researchers mash tissue into a tube and sequence everything at once, they get an average. That average can hide the very signals that matter most, like a small pocket of drug-resistant cancer cells or a cluster of immune cells that failed to attack a tumor. Spatial transcriptomics technologies solve this by profiling cells within their tissue context, making it possible to see which cell types neighbor each other, how their gene activity changes depending on that neighborhood, and how molecular signals ripple across a tissue’s architecture.1PubMed. Single-cell and spatial transcriptomics: Bridging current technologies with long-read sequencing
The ability to pin hundreds or even thousands of different molecules to precise spots in a tissue section is what makes these tools so powerful. Rather than inferring that two cell types probably interact because their gene profiles suggest it, researchers can see the cells sitting side by side and watch the molecular conversation unfold.2Nature Reviews Nephrology. Spatial transcriptomics in health and disease
The Main Technology Families
There is no single “spatial sequencing machine.” Instead, a cluster of competing approaches has emerged, each with different strengths. They broadly fall into three camps: array-based capture, imaging-based detection, and region-of-interest profiling.
Array-Based Capture
In array-based methods, a thin tissue section is placed on a glass slide covered in tiny spots, each carrying a unique molecular barcode. As RNA leaks out of the tissue and lands on the slide, each barcode tags the RNA with a spatial address. Researchers then sequence the tagged RNA and computationally map it back to the tissue. One version of this approach packs barcoded beads into microsized wells on a slide, then decodes each bead’s position through sequential rounds of labeling, achieving high spatial density.3bioRxiv. High-density spatial transcriptomics arrays for in situ tissue profiling The advantage is genome-wide coverage: in principle, every gene expressed in the tissue can be detected. The trade-off is resolution. Because RNA diffuses slightly before it lands on a barcode, the effective resolution can blur, and a single “spot” may capture RNA from more than one cell.
Imaging-Based Detection
Imaging methods take the opposite approach. Instead of capturing RNA on a slide, they label RNA molecules directly inside the intact tissue using fluorescent probes and then photograph them. MERFISH, one of the best-known imaging platforms, uses error-correcting encoding schemes to identify hundreds to thousands of different RNA species in individual cells by cycling through rounds of fluorescent labeling and imaging.4PubMed Central. Spatially resolved, highly multiplexed RNA profiling in single cells Because each RNA molecule is visualized one at a time, MERFISH naturally achieves single-molecule, subcellular resolution. You can see not just that a gene is active in a cell, but where in the cell the RNA is located, whether clustered near the nucleus or transported toward the cell’s edge.5PubMed Central. Spatial transcriptome profiling by MERFISH reveals subcellular RNA compartmentalization and cell cycle-dependent gene expression The downside is that imaging methods typically target a pre-selected panel of genes rather than the entire genome, so researchers must choose in advance which genes to look for.
In Situ Sequencing
A third family of methods reads out genetic sequences directly inside the tissue. Short probe molecules bind to target RNAs, get enzymatically circularized, and then are amplified on the spot through a process called rolling circle amplification. The amplified signal is bright enough to be read through sequential rounds of fluorescent labeling, effectively sequencing a short barcode in place. This approach has been used, for example, to map the spatial expression patterns of dozens of genes in mouse kidneys, comparing healthy and diabetic tissue.6PubMed. Efficient Spatial Gene Expression Profiling Using Split-Probe Ligation and Rolling Circle Amplification A related method, BaristaSeq, adapted this probe-based sequencing strategy specifically for reading cellular barcodes, allowing researchers to trace neuronal wiring patterns in the brain.7PubMed Central. Efficient in situ barcode sequencing using padlock probe-based BaristaSeq
Region-of-Interest Profiling
Some platforms skip the goal of mapping every cell in a tissue and instead let researchers pick specific regions to profile. Digital Spatial Profiling uses antibodies or RNA probes tagged with light-sensitive barcodes. A researcher views the tissue under a microscope, selects a region of interest, and a device shines precisely patterned UV light onto that region to release the barcodes, which are then collected and sequenced. The light pattern can be shaped to cover a hand-drawn area, a specific cell type identified by a stain, or even a spatial gradient radiating outward from a feature like a tumor boundary.8Nature Biotechnology. Multiplex digital spatial profiling of proteins and RNA in fixed tissue This flexibility is especially attractive for clinical samples where the question is not “what is everywhere in this tissue” but “what is happening right here, at the tumor’s invasive front.”9PubMed. High-Plex Spatially Resolved RNA and Protein Detection Using Digital Spatial Profiling
Computational Challenges That Shape the Data
Generating spatial data is only half the problem. Making sense of it requires specialized computational tools, and two challenges in particular have driven a wave of algorithm development.
Figuring Out What Is in Each Spot
In array-based methods, a single capture spot often contains RNA from multiple cells. Researchers use deconvolution algorithms to estimate the mixture of cell types in each spot by comparing the spot’s gene expression profile against a reference catalog of known cell types built from single-cell sequencing. Tools like SpatialDecon and CARD take different approaches to this problem. SpatialDecon introduced several refinements to standard gene-expression deconvolution for spatial data.10Nature Communications. Advances in mixed cell deconvolution enable quantification of cell types in spatial transcriptomic data CARD goes a step further by incorporating the spatial relationships between neighboring spots, reasoning that if a spot’s neighbors are mostly immune cells, the spot itself is more likely to contain immune cells too.11Nature Biotechnology. Spatially informed cell-type deconvolution for spatial transcriptomics Another tool, DestVI, pushes beyond simple cell-type labels to identify continuous variation within a cell type, like a gradient of activation states in immune cells.12PubMed Central. DestVI identifies continuums of cell types in spatial transcriptomics data
Drawing Cell Boundaries
Imaging-based methods face a different bottleneck: figuring out which detected RNA molecules belong to which cell. This is the cell segmentation problem. Cells in tissue are irregularly shaped and tightly packed, and a transcript floating near the boundary between two cells could belong to either one. Baysor, one segmentation tool, optimizes cell boundaries in two or three dimensions by jointly considering the transcriptional identity of nearby molecules and any available cell-shape information from staining images. It can even segment cells using transcript positions alone, without any staining at all.13Nature Biotechnology. Cell segmentation in imaging-based spatial transcriptomics A newer tool called Segger frames the problem differently, treating each transcript as a node in a graph and using a neural network to predict which cell each transcript belongs to.14PubMed Central. Segger: Fast and accurate cell segmentation of imaging-based spatial transcriptomics data Getting segmentation wrong cascades through every downstream analysis, so this is an active and consequential area of development.
Mapping Tumors in Their Natural Habitat
Cancer research has been one of the most enthusiastic adopters of spatial sequencing, and for good reason. Tumors are ecosystems, not just clusters of mutated cells. The mix of cancer cells, immune cells, blood vessels, and connective tissue surrounding them, collectively called the tumor microenvironment, heavily influences whether a patient responds to treatment. Spatial transcriptomics allows researchers to examine distinct tumor sub-regions: the core, the invasive front, immune-infiltrated zones, and even rare structures like tertiary lymphoid structures where immune cells organize as if they were in a lymph node.15Cell Death Discovery. Spatial transcriptomics in cancer research: insights into tumorigenesis, diagnosis and therapeutics
A pan-cancer spatial analysis across multiple tumor types recently identified recurring microenvironment “niches,” defined by which cell types co-localize. Some niches featured macrophages sitting directly among tumor cells; others showed lymphocytes infiltrating the tumor; still others resembled organized immune structures with B cells, dendritic cells, and T cells aggregating together in patterns reminiscent of lymph nodes.16Cell Reports Medicine. Pan-cancer spatial transcriptomics map reveals the conserved niches and clinical implications of the tumor spatial microenvironment These patterns appear conserved across cancer types, suggesting shared organizational principles that could eventually guide treatment decisions. Understanding not just which immune cells are present but where they are relative to the tumor, whether they are excluded at the boundary or actively infiltrating, is increasingly recognized as critical for predicting immunotherapy response.17PubMed Central. Unveiling the role of spatial transcriptomics in the analysis of the tumor immune microenvironment (Review)
Charting the Brain
Neuroscience has proved to be an equally fertile ground. The brain contains an astonishing diversity of cell types arranged in precise spatial patterns, and traditional methods of classifying neurons by their gene expression profiles lose the spatial relationships that define brain circuits. A whole-mouse-brain atlas built using Stereo-seq combined with single-nucleus sequencing cataloged over four million cells across 308 cell clusters, with each cluster mapped to its physical position in the brain, covering nearly 30,000 genes.18Neuron. Single-cell spatial transcriptomic atlas of the whole mouse brain
Human brain studies have followed. Using MERFISH augmented with deep-learning-based cell identification, researchers created an atlas of the developing human fetal cortex spanning more than 18 million single cells across multiple brain areas and developmental time points. The data revealed how cortical layers and specialized brain regions acquire their molecular identities during fetal development.19PubMed Central. Spatial transcriptomics reveals human cortical layer and area specification
Disease applications are equally compelling. In Alzheimer’s research, spatial transcriptomics has been used to profile the gene expression changes that occur in the immediate vicinity of amyloid plaques, the protein deposits that are a hallmark of the disease. One study mapped these changes within a 100-micrometer radius around plaques in a mouse model, identifying a dysregulated cellular network that traditional bulk approaches would have averaged away.20PubMed. Spatial Transcriptomics and In Situ Sequencing to Study Alzheimer’s Disease Follow-up work at higher resolution revealed that glial cells near plaques disrupt neuronal signaling in specific, spatially patterned ways, with the disruption fading as distance from the plaque increased.21Cell Reports. High-resolution spatial transcriptomics reveals glial-induced disruption of neuronal signaling in the amyloid plaque niche
Seeing Cell Conversations in Place
Cells do not operate in isolation. They send and receive molecular signals through ligand-receptor pairs: one cell releases a signaling molecule (the ligand) and a neighboring cell detects it with a matching receptor. Traditional bioinformatics tools infer these interactions from gene expression data alone, guessing which cells might be talking based on which ligands and receptors they express. Spatial data transforms this guesswork into direct observation. SpatialDM, for instance, uses a statistical approach to identify ligand-receptor pairs that are spatially co-expressed, meaning the ligand and receptor are not just both present in the tissue but are found in the same neighborhoods. Applied to melanoma and intestinal tissues, it identified context-specific communication patterns and differential interactions between healthy and diseased conditions.22PubMed Central. SpatialDM for rapid identification of spatially co-expressed ligand–receptor and revealing cell–cell communication patterns
Embryos, Microbes, and Multi-omics
Developmental biology has embraced spatial sequencing to track how an organism builds itself from a single cell. A recent study mapped the expression of 495 genes across entire zebrafish embryos at subcellular resolution, from the earliest stages of body-plan formation through organ development, and paired this with chromatin accessibility data to connect gene activity to the regulatory elements controlling it.23PubMed. Whole-embryo spatial transcriptomics at subcellular resolution from gastrulation to organogenesis
The intersection of spatial sequencing with microbiology is also opening new doors. In Crohn’s disease, researchers developed a method to simultaneously identify bacterial species and measure host gene expression in the same tissue section, revealing how specific bacteria alter the transcriptional behavior of nearby intestinal cells.24PubMed Central. Spatial host-microbiome profiling demonstrates bacterial-associated host transcriptional alterations in pediatric ileal Crohn’s disease A complementary framework called MicroCart enables simultaneous probing of host and microbiome across multiple spatial modalities within the intestinal tissue microenvironment.25Nature Communications. A multi-omics spatial framework for host-microbiome dissection within the intestinal tissue microenvironment
Meanwhile, the field is rapidly moving beyond RNA alone. Technologies now exist to co-profile chromatin accessibility and gene expression on the same tissue section, linking which stretches of DNA are open and available for reading with which genes are actually being read.26PubMed Central. Spatial epigenome-transcriptome co-profiling of mammalian tissues Others simultaneously measure proteins and RNA at every spot, combining antibody-based protein detection with transcriptome-wide RNA capture on the same array.27PubMed Central. Integration of whole transcriptome spatial profiling with protein markers One method, MiP-seq, goes further still, detecting DNA, RNA, proteins, and other biomolecules simultaneously at subcellular resolution.28Nature Biomedical Engineering. Spatial multi-omics at subcellular resolution via high-throughput in situ pairwise sequencing These multi-omic approaches are converging toward a future where a single tissue section yields a comprehensive molecular portrait of every cell, from its DNA state through its RNA activity to its protein output.
Getting Spatial Data from Clinical Samples
One of the most practically important advances in spatial sequencing is its growing compatibility with the tissue samples that already sit in hospital archives. The standard way to preserve tissue for pathology is to fix it in formalin and embed it in paraffin wax, producing what are called FFPE blocks. These are the most abundant archivable specimens in clinical tissue banks worldwide, but the fixation process fragments RNA, which made these samples largely incompatible with earlier sequencing methods. New approaches that barcode RNA directly inside the tissue, without needing to extract it first, have opened up FFPE samples to spatial profiling.29bioRxiv. Spatial transcriptome sequencing of FFPE tissues at cellular level A more recent technology, spRandom-seq, uses random primers to capture total RNA spatially from FFPE tissue, and can even detect bacterial and viral RNA alongside host transcripts in the same section.30PubMed Central. Spatial Total RNA Sequencing of Formalin-Fixed Paraffin-Embedded Tissue by spRandom-seq
This matters enormously for clinical translation. Hospitals have decades’ worth of FFPE blocks linked to patient outcomes, drug responses, and survival data. If spatial sequencing can reliably work on those archived samples, it unlocks retrospective studies that would be impossible to run prospectively. Several spatial methods have now achieved FFPE compatibility, and there is growing discussion about whether these technologies could eventually be incorporated into routine pathology practice.31PubMed. Spatial transcriptomics and the anatomical pathologist: Molecular meets morphology
Benchmarking and Practical Trade-offs
With so many competing platforms, researchers face genuine decisions about which technology to use for a given question. A systematic comparison of 11 sequencing-based spatial transcriptomics methods highlighted some important variables. Molecular diffusion, the tendency of RNA to spread slightly before being captured, varied across methods and tissue types, significantly affecting the effective resolution. The study also found that spatial data requires substantially more sequencing to reach saturation compared to standard single-cell data, and that platforms differed in their capture efficiency. Stereo-seq, Slide-tag, and Visium (probe-based) showed strong capture efficiency at raw sequencing depth, while different platforms led when efficiency was normalized.32Nature Methods. Systematic comparison of sequencing-based spatial transcriptomic methods Cost also varies considerably. Open-ST, for example, was reported to require the least sequencing depth to obtain equivalent transcriptomic information, making it more accessible for labs on a budget.33Cell. High-resolution spatial mapping of single cells in 3D through Open-ST
No single platform wins on every axis. Imaging-based methods give the best resolution but require pre-selecting genes. Array-based methods are genome-wide but sacrifice resolution. Region-of-interest tools are flexible and work well on clinical tissue but do not produce whole-tissue maps. Choosing among them depends on whether a researcher needs subcellular detail, whole-transcriptome breadth, compatibility with archival tissue, or high throughput at low cost.
Building Tissues in Three Dimensions
Most spatial sequencing experiments produce two-dimensional maps from thin tissue slices. But organs are three-dimensional, and important features, like a blood vessel branching through a tumor or the layered architecture of the brain, span many slices. Reconstructing 3D tissue volumes from serial sections is a growing computational frontier. STAIR, a tool that uses a graph-based neural network with multi-level attention mechanisms, can align and integrate multiple slices into a unified 3D atlas, including reconstruction along the depth axis between parallel sections.34PubMed Central. Spatial transcriptomic alignment, integration, and 3D reconstruction by STAIR Other reconstruction methods have been benchmarked on datasets as large as 129 serial brain slices containing over nine million cells, stitching them into volumetric models that reveal spatial patterns invisible in any single two-dimensional section.35Nature Communications. Network model for alignment, stitching and slice-to-volume 3D reconstruction of large-scale spatially resolved slices
Adapting the Technology for Plants
Nearly all spatial sequencing platforms were developed for mammalian tissues, and plant biologists face real obstacles in adopting them. Plant cells have rigid cell walls, large fluid-filled vacuoles, and high levels of autofluorescence that interfere with fluorescent imaging.36bioRxiv. Plant-Compatible Xenium In Situ Sequencing: Optimised Protocol for Spatial Transcriptomics in Medicago truncatula Roots and Nodules The cell walls block probe access, the vacuoles reduce the density of RNA-containing cytoplasm per cell, and the autofluorescence creates background noise that complicates transcript detection. These are not insurmountable problems, but they require protocol modifications that are still being worked out. Computational pipelines also need adjusting, since the size variability of plant cells is much greater than in animal tissues, and standard segmentation algorithms tuned for compact mammalian cells can perform poorly.37Trends in Plant Science. Plant spatial multi-omics: from experimental strategies to data analysis workflows Early results from optimized protocols for species like the legume Medicago truncatula suggest the technology is transferable, but plant spatial transcriptomics remains years behind its mammalian counterpart in maturity and standardization.