Understanding Your Biology, One Cell at a time
Single-cell and spatial transcriptomics have fundamentally changed how we study complex biological systems. Where bulk RNA sequencing reports the average gene expression of a tissue, obscuring the contributions of individual cell types, rare populations, and transitional states. Single-cell and spatial methods resolve gene expression at cellular and subcellular resolution. This page provides an overview of what these technologies can reveal and how modern analytical approaches extract biological meaning from the data.
Single-Cell RNA Sequencing
Single-cell RNA sequencing (scRNA-seq) measures the transcriptome of individual cells, typically thousands to tens of thousands of cells per sample. Each cell is captured, its mRNA reverse-transcribed, and a sequencing library constructed that carries a unique cellular barcode. After sequencing and alignment, the result is a gene expression matrix, genes by cells, that serves as the starting point for downstream analysis.
What single-cell data reveals
Cell type composition. Unsupervised clustering of single-cell data groups cells by transcriptional similarity, enabling identification of major and minor cell populations without prior knowledge of marker genes. Reference atlases and automated annotation tools (such as CellTypist) allow systematic assignment of cell identities across tissues and species.
Rare and transitional cell populations. Bulk sequencing averages away rare cell types that may represent fewer than 1% of a tissue but play disproportionately important biological roles such as tumor-initiating cells, tissue-resident immune subsets, and senescent cells. Single-cell resolution makes these populations accessible.
Differential gene expression at cell-type resolution. Rather than asking “what genes change between treated and untreated tissue,” single-cell data allows the question to be asked within each cell type independently. A treatment effect visible only in one cell population invisible in bulk becomes apparent.
Cellular trajectories and developmental dynamics. Pseudotime analysis (Monocle, scVelo) orders cells along inferred developmental or activation trajectories, identifying gene regulatory programs that drive differentiation, activation, or senescence. RNA velocity extends this by estimating the future transcriptional state of each cell from the ratio of spliced to unspliced transcripts.
Cell-cell communication. Ligand-receptor analysis (CellChat, NicheNet) systematically identifies signaling interactions between cell populations, revealing which cells are sending and receiving molecular signals and through which pathways.
Gene regulatory networks. Methods such as SCENIC reconstruct transcription factor regulons, the transcription factors active in each cell and their downstream target genes, providing a mechanistic layer beneath the differential expression results.
Fresh vs. Fixed Single-Cell
Single-cell platforms fall into two fundamentally different chemistries, and the distinction matters for experimental design.
Fresh cell capture (3′ and 5′ gene expression) isolates intact cells or nuclei from fresh or fresh-frozen tissue and captures mRNA via the poly-A tail. Because capture relies on polyadenylation rather than probe hybridization, this approach is species-agnostic and works on any organism that produces polyadenylated mRNA. Fresh single-cell is the default choice for most experiments and supports the full range of multiomic add-ons: cell surface protein profiling (CITE-seq), TCR/BCR immune repertoire sequencing, CRISPR perturbation screens, and ATAC-seq chromatin accessibility.
Fixed cell capture (Chromium Fixed RNA, GEM-X Flex) uses probe hybridization to capture gene expression from formalin-fixed cells or nuclei. This enables processing of FFPE tissue and allows samples to be fixed at the point of collection and batched for later processing, reducing technical variation across samples. However, because detection relies on pre-designed probe sets, this approach is currently limited to human and mouse. Custom probe design is available for genes outside the standard panel.
Single-nucleus RNA sequencing (snRNA-seq) is often the method of choice for tissues that are difficult to dissociate such as brain, heart, or skeletal muscle, where intact single-cell isolation would cause stress-induced transcriptional artifacts or yield primarily debris. Nuclei are more robust to mechanical and enzymatic dissociation and can be isolated from fresh-frozen or FFPE tissue.
Spatial Transcriptomics
Single-cell sequencing dissociates tissue into individual cells, losing the spatial relationships between them. Spatial transcriptomics preserves tissue architecture while measuring gene expression, enabling questions that single-cell data alone cannot answer: where in the tissue does this cell type reside? Which cell populations are in physical proximity? How does gene expression change across a tissue gradient?
VisiumHD
10x Genomics VisiumHD captures whole-transcriptome gene expression on a continuous 2µm × 2µm barcoded grid laid over an intact tissue section. Unlike the original Visium platform (which captured RNA from 55um diameter spots, VisiumHD achieves single-cell-scale resolution across the entire tissue section. A single capture area covers 6.5mm × 6.5mm of tissue (small format) or 11mm x 11mm (large format), sufficient for most biopsy and tissue section sizes.
VisiumHD uses probe-based chemistry, which requires tissue fixation. Like GEM-X Flex, it is currently limited to human and mouse probe sets, though probe homology analysis can inform the expected detection rate for closely related species. 3DG has validated VisiumHD on non-human primate tissue using the human probeset, achieving ~90% predicted gene detection.
Cell segmentation, the assignment of individual bins to specific cells, is performed computationally using nuclear staining and isotropic radius expansion from detected nuclei. This approach is an approximation; neither VisiumHD nor Xenium directly detects cell membrane boundaries. Segmentation accuracy is a meaningful source of technical noise in spatial data, and parameter choices (particularly the isotropic expansion radius) significantly affect the fraction of transcripts assigned to the correct cell. 3DG has characterized this in detail in our technical whitepaper on segmentation error.
Xenium In Situ
Xenium uses fluorescence in situ hybridization (FISH) with padlock probes and rolling circle amplification to detect individual RNA molecules in intact tissue sections. Rather than sequencing, Xenium reads out gene expression through iterative rounds of fluorescence imaging and decoding. This approach provides single-molecule spatial resolution and supports larger tissue areas (up to 10mm × 20mm) than VisiumHD.
The tradeoff is gene coverage: current Xenium panels cover up to 5,000 genes, compared to the ~18,000-gene whole-transcriptome coverage of VisiumHD. Xenium supports custom probe design for species beyond human and mouse.
VisiumHD and Xenium as Complementary Platforms
VisiumHD and Xenium are complementary, not competing, technologies. VisiumHD provides unbiased whole-transcriptome coverage across the tissue, ideal for discovery and hypothesis generation. Xenium provides single-molecule resolution validation of specific targets at larger tissue scale. A common workflow uses VisiumHD to identify cell types and spatially variable genes of interest, then validates and extends findings with a targeted Xenium panel, potentially across larger cohorts or tissue areas.
From Data to Biology
The biological value of single-cell or spatial data is realized through computational analysis. Below we outline the analytical approaches most commonly applied to single-cell and spatial data.
Standard pipeline
Raw sequencing reads are aligned to the reference genome and quantified using Cell Ranger (10x Genomics) or equivalent pipelines. Quality control removes low-quality cells, doublets (DoubletFinder), and ambient RNA contamination (SoupX, CellBender). Data are normalized, dimensionality-reduced (PCA, UMAP), and clustered. Cell types are annotated using marker genes and reference atlases. Go deeper.
Differential expression and pathway analysis
Pseudobulk differential expression (DESeq2, edgeR) identifies genes that change between conditions, time points, or disease states within each cell type identified via UMAP or TSNE clustering. Gene set enrichment analysis (GSEA, fgsea) maps differentially expressed genes to biological pathways, providing mechanistic interpretation of transcriptional changes. Go deeper.
Rare and specialized cell types
Many experiments are designed specifically to find cells that standard clustering misses — rare progenitor populations, transient differentiation intermediates, and stress-induced states that appear briefly before resolving. Specialized methods evaluate multi-marker combinations, rank cell types by condition-specific separability, and correct for doublets and ambient RNA contamination that can masquerade as rare populations. Go deeper.
Tumor biology
Single-cell and spatial transcriptomics are particularly powerful in oncology, where resolving tumor, immune, and stromal compartments at single-cell resolution is essential for understanding treatment response and resistance. Go deeper.
Clinical and translational applications
Clinical and translational studies impose additional analytical requirements. Multi-sample designs require pseudobulk differential expression to avoid inflated false positive rates, and batch correction across collection sites and time points is essential for cohort-level comparisons. Reference-based cell type annotation enables reproducible classification across studies and institutions. Go deeper.
Trajectory and RNA velocity analysis
Pseudotime methods order cells along differentiation or activation trajectories, enabling reconstruction of dynamic biological processes from static snapshots. RNA velocity (scVelo, UniTVelo) estimates the direction and speed of transcriptional change in each cell, providing a time-resolved view of cellular dynamics. Go deeper.
Cell-cell communication
Ligand-receptor databases (CellChat, CellPhoneDB, NicheNet) systematically identify intercellular signaling interactions. Spatial data adds a critical constraint as only cell types in physical proximity can engage in paracrine signaling, enabling biologically grounded communication analysis. Go deeper
Gene regulatory network inference
SCENIC and related tools reconstruct transcription factor regulons from single-cell data, identifying the master regulators of cell identity and state transitions. This provides a mechanistic layer that connects transcriptional changes to upstream regulatory logic. Go deeper.
Multi-modal and cross-platform integration
When single-cell and spatial data are collected from matched samples, integration methods (Seurat, Harmony, scVI) enable cell type assignments from single-cell data to be projected onto spatial maps, combining the resolution of single-cell profiling with the spatial context of tissue architecture. Integration across batches, platforms, and species requires careful normalization and batch correction. Go deeper.
Applying These Approaches
The technologies and analytical methods described here are tools for answering biological questions. The right experimental design depends on the question being asked, the tissue type, the available sample material, and the downstream analytical goals.
3DG applies these approaches across a range of biological systems and disease areas. If you’re considering single-cell or spatial transcriptomics to address your research questions, we’re happy to discuss experimental design, platform selection, and what you might expect from the data.