- Integration of single-cell and spatial data is an important approach in modern bioinformatics that combines detailed molecular information from individual cells with their physical locations and organization within tissues. Single-cell technologies, particularly single-cell RNA sequencing (scRNA-seq), reveal gene expression patterns, cellular identities, functional states, and molecular heterogeneity at individual-cell resolution. Spatial technologies, including spatial transcriptomics and imaging-based molecular profiling, preserve information about where cells and their gene-expression patterns occur within tissues. Integrating these complementary data types provides a more complete understanding of tissue organization, cellular diversity, molecular interactions, and biological processes.
- Single-cell RNA sequencing provides detailed information about gene expression in individual cells. It can identify cell types, distinguish cellular subpopulations, characterize developmental and functional states, and reveal rare cell populations that may be difficult to detect using bulk RNA sequencing. However, conventional scRNA-seq generally requires tissue dissociation, which removes the original spatial arrangement of cells. Consequently, it cannot directly establish where a particular cell population was located or which other cell types were physically nearby before tissue processing.
- Spatial transcriptomics addresses this limitation by measuring gene expression while retaining information about tissue coordinates. Different platforms provide different levels of spatial resolution. Spot-based technologies such as 10x Visium and Slide-seq measure expression within defined tissue locations that may contain multiple cells, while high-resolution technologies such as Visium HD provide finer spatial measurements. Imaging-based platforms such as Xenium, MERFISH, seqFISH, and CosMx can measure selected or extensive sets of RNA molecules within individual cells or at subcellular resolution, depending on the platform. Spatial proteomics and multiplexed imaging provide additional information about protein abundance and localization.
- The integration of single-cell and spatial data combines the strengths of these approaches. Single-cell data provide detailed cellular expression profiles, while spatial data reveal the tissue locations and organization of cells. Together, they help determine which cell types occupy particular tissue regions, how cellular states vary across anatomical structures, and how neighboring populations may interact. For example, single-cell sequencing may identify several macrophage subpopulations in a tumor, while spatial data can help determine whether these populations are located near cancer cells, blood vessels, or immune-rich regions.
- One of the most common integration strategies is reference-based cell-type mapping. In this approach, a single-cell RNA-seq dataset is used as a reference containing annotated cell types and their gene-expression profiles. Spatial transcriptomics data are treated as a query dataset in which the cellular identities or proportions are initially unknown or incompletely characterized. Computational algorithms compare the expression patterns in the spatial data with reference profiles and estimate which cell types are present in each spatial location. This is particularly useful for spot-based spatial transcriptomics, where each measurement may represent a mixture of several cells.
- Cell-type deconvolution is closely related to reference-based mapping. It estimates the proportions of different cell types contributing to a spatial measurement. For example, a tissue spot may contain epithelial cells, macrophages, T lymphocytes, fibroblasts, and endothelial cells. Deconvolution algorithms use reference expression profiles to estimate the contribution of each population. Common methods include RCTD, cell2location, SPOTlight, stereoscope, and DestVI. These approaches differ in their statistical assumptions, treatment of technical variation, and ability to model cellular heterogeneity. Tangram provides another approach for mapping single-cell profiles to spatial locations by aligning molecular expression patterns between single-cell and spatial datasets.
- Cell2location is a probabilistic framework designed to estimate cell-type abundance in spatial transcriptomics data using single-cell reference information. RCTD uses statistical modeling to account for mixtures of cell types in spatial measurements. SPOTlight uses non-negative matrix factorization and marker information to estimate cell-type proportions, while stereoscope uses probabilistic modeling to deconvolve spatial measurements. DestVI models both cell-type abundance and variation in cell states across spatial locations. These methods can provide complementary results, and method selection should depend on the spatial platform, reference dataset, cell-type diversity, and biological question.
- Another integration strategy is spatial cell-type mapping, in which individual cells identified through imaging or high-resolution spatial transcriptomics are assigned molecular identities using single-cell reference data. This can help distinguish cellular populations that share similar morphological characteristics but have different molecular functions. Tools such as Seurat, scmap, SingleR, CellTypist, Azimuth, and scArches support cell annotation, reference mapping, or data transfer, although their intended inputs and outputs differ. Some are primarily designed for single-cell reference annotation and transfer rather than direct spatial deconvolution, so their suitability depends on the spatial data structure and analysis objective.
- Data integration may also involve combining gene-expression matrices from different technologies. Single-cell and spatial datasets often differ in sequencing depth, gene detection, measurement resolution, technical noise, and gene coverage. Before integration, researchers perform quality control, normalization, feature selection, and batch-effect assessment. Common preprocessing methods include Seurat and Scanpy for single-cell analysis, while spatial workflows may use Space Ranger, Xenium Ranger, and platform-specific processing software. Spatial data can be organized and analyzed using frameworks such as SpatialExperiment in R and AnnData in Python.
- Batch correction and data harmonization are important because single-cell and spatial datasets may come from different experiments, platforms, laboratories, tissue preparations, or biological conditions. Differences in technology can produce expression patterns that reflect technical variation rather than genuine biology. Integration approaches such as Seurat integration, Harmony, fastMNN, and scVI can help align compatible datasets and reduce unwanted technical variation. However, integration should not eliminate genuine biological differences between cell populations, disease states, or tissue regions. Researchers should therefore evaluate whether corrected data preserve known cell-type markers, biological variation, and spatial patterns.
- Spatial information can also improve the interpretation of cell states identified through single-cell analysis. A single-cell dataset may reveal different activation states of T lymphocytes, macrophages, fibroblasts, or epithelial cells, but it may not indicate where these states occur within a tissue. Spatial mapping can determine whether particular cell states are concentrated near blood vessels, tumor boundaries, inflammatory regions, or specialized anatomical structures. This can reveal tissue-specific functions and help distinguish cellular populations that appear similar in conventional single-cell clustering.
- The integration of single-cell and spatial data is also valuable for identifying spatially organized cellular neighborhoods. Once cell types or cell states have been mapped to tissue locations, spatial graph methods can identify neighboring populations and quantify their local composition. Tools such as Squidpy and Giotto support spatial graph construction, neighborhood enrichment, co-occurrence analysis, and other spatial statistical methods. These analyses can reveal specialized tissue compartments, immune-cell aggregates, tumor-associated niches, and developmental regions.
- Another important application is the investigation of spatial cell–cell communication. Single-cell data can identify ligand-producing and receptor-expressing cell populations, while spatial data reveal whether these populations occur near one another. Communication tools such as CellChat, CellPhoneDB, NicheNet, and LIANA can be combined with spatial coordinates or neighborhood information to investigate candidate signaling relationships. This integration helps prioritize potential interactions between cells in biologically relevant locations. Nevertheless, gene expression and spatial proximity alone do not prove that a signaling pathway is active; protein-level and functional validation may be necessary.
- Integration can also be extended to spatially variable gene analysis. Spatially variable genes are genes whose expression changes according to tissue location or spatial structure. Once cell types are mapped, researchers can investigate whether specific genes show spatial patterns within particular cellular populations or anatomical regions. Methods such as SpatialDE, SPARK, and SPARK-X can identify spatially structured expression patterns, while Squidpy and Giotto provide additional spatial analysis capabilities. Combining cell-type information with spatial gene-expression patterns can help distinguish changes caused by differences in cellular composition from changes occurring within the same cell type.
- Several computational frameworks support the integration of single-cell and spatial data. Seurat provides tools for single-cell analysis, reference mapping, data transfer, and spatial transcriptomics workflows. Scanpy offers a Python-based framework for single-cell analysis and can be combined with Squidpy for spatial analysis. Giotto provides tools for spatial transcriptomics, cellular organization, and spatial interaction analysis. SpatialExperiment supports standardized handling of spatially resolved expression data in R. scVI and scArches provide deep-learning-based approaches for data integration and reference mapping in suitable single-cell and spatial workflows. Tangram, cell2location, RCTD, SPOTlight, stereoscope, and DestVI are particularly relevant to mapping and deconvolution tasks.
- A typical integration workflow begins with the collection of compatible single-cell and spatial datasets from the same tissue type, biological condition, or experimental system. Single-cell data undergo quality control, normalization, dimensionality reduction, clustering, and cell-type annotation. Spatial data are processed to identify tissue regions, spatial coordinates, gene-expression measurements, and, where applicable, individual cells. The datasets are then harmonized to address differences in gene identifiers, feature coverage, normalization, and technical effects. A suitable reference-mapping or deconvolution method is applied to estimate cell identities or cell-type proportions in spatial locations. The resulting maps can then be used for spatial clustering, neighborhood analysis, spatially variable gene detection, and cell–cell communication analysis.
- Visualization is an essential part of this process. Spatial maps can display the distribution of predicted cell types across tissue sections, while feature plots show the expression of selected genes in their spatial context. Heatmaps can compare cell-type abundance across anatomical regions, and interaction networks can represent candidate communication between cell populations. Tools such as Seurat, Scanpy, Squidpy, Giotto, QuPath, and Fiji/ImageJ support different aspects of data visualization and tissue-image interpretation. Histological images, including hematoxylin and eosin staining, immunohistochemistry, and immunofluorescence, can provide additional anatomical and cellular context.
- Integration of single-cell and spatial data has important applications in cancer research. Single-cell sequencing can identify tumor cells, immune populations, fibroblast subtypes, and endothelial cell states, while spatial analysis reveals their locations within tumor regions. This helps characterize tumor boundaries, immune-infiltrated areas, immunosuppressive niches, stromal compartments, and potential sites of invasion. In neuroscience, integrated analysis can map neuronal and glial cell populations to brain regions and investigate their local molecular environments. In developmental biology, it can reveal how cell differentiation and signaling occur across developing tissues. In immunology and infection biology, it can identify spatially organized immune responses and interactions between host cells and pathogens.
- The approach is also valuable in veterinary medicine, animal genomics, and animal breeding research. It can help investigate cellular diversity and tissue organization in livestock, companion animals, and other species. Examples include mapping immune-cell populations in infected tissues, identifying cellular environments associated with disease resistance, studying muscle-cell development, characterizing reproductive tissues, and examining inflammatory responses. Comparative integration across breeds or species may help identify conserved cellular programs and tissue-specific molecular differences associated with health, adaptation, and production traits. However, species-specific reference datasets and reliable gene-annotation resources are important for accurate mapping.
- Despite its advantages, integration of single-cell and spatial data presents several challenges. Single-cell and spatial technologies differ in gene coverage, sequencing depth, resolution, sensitivity, and technical noise. A reference dataset may not represent every cell type or cell state present in the spatial sample. Differences in tissue preparation, dissociation, cell viability, and experimental conditions can also affect gene-expression profiles. Spot-based spatial platforms introduce uncertainty because multiple cells contribute to a single measurement, while imaging-based platforms may measure only a selected panel of genes. Incorrect cell-type annotations or inappropriate reference datasets can lead to misleading spatial maps.
- Another challenge is distinguishing true biological relationships from computational predictions. A cell type assigned to a spatial region may be supported by a statistical model but still require independent validation. Deconvolution estimates depend on reference quality and model assumptions, and different algorithms may produce different cell-abundance estimates. Researchers should compare suitable methods, evaluate marker-gene consistency, perform sensitivity analyses, and use biological replicates. Histological examination, immunofluorescence, immunohistochemistry, RNA in situ hybridization, and spatial proteomics can provide independent validation of important findings.
- Overall, integration of single-cell and spatial data is a powerful strategy in bioinformatics because it connects detailed cellular molecular profiles with the spatial organization of tissues. It enables researchers to identify cell types, estimate cellular composition, map cell states, investigate tissue neighborhoods, detect spatial gene-expression patterns, and study potential cell–cell communication. As single-cell sequencing, spatial transcriptomics, imaging technologies, and computational methods continue to develop, their integration will provide increasingly detailed insights into tissue biology, disease mechanisms, development, veterinary medicine, and animal genomics.