Single-Cell and Spatial Biology Services: From scRNA-Seq to Spatial Transcriptomics for Precision Cell Analysis

Single-cell RNA sequencing resolves gene expression at individual-cell resolution; spatial transcriptomics maps those expression patterns onto native tissue architecture. Together they answer what cell types exist, what states they occupy, and where they reside. This guide covers platform selection, experimental design trade-offs, and how to match biological questions to the right technology constellation.

Figure 1: Single-Cell & Spatial Biology Technology Landscape — scRNA-seq, Spatial, and Multi-Omics Platforms Overview

From Bulk Averages to Cellular Resolution

The central limitation of bulk RNA sequencing is not sensitivity — it is averaging. A bulk RNA-seq sample from a tumor contains transcripts from malignant cells, infiltrating T cells, stromal fibroblasts, endothelial cells, and macrophages, each expressing a fundamentally different gene program. The bulk measurement collapses this heterogeneity into a single weighted average. Differential expression analysis between tumor and adjacent normal tissue may identify hundreds of statistically significant genes, but cannot tell you whether the IL6 upregulation comes from the cancer cells or the infiltrating immune cells.

scRNA-seq solves this by encapsulating individual cells in nanoliter-scale droplets, tagging each cell's transcripts with a unique barcode, and sequencing them separately. A single 10x Genomics Chromium run profiles 500 to 20,000 cells in parallel, generating a matrix where rows are genes, columns are cells, and each entry is the number of transcript molecules detected (1). The result is not one expression profile per sample, but thousands — each a window into a single cell's transcriptomic state.

The practical implications are profound. A scRNA-seq study of non-small cell lung cancer that profiled 50,000 cells identified 12 distinct immune and stromal populations whose relative abundances predicted patient prognosis — information invisible in the bulk average. Studies of checkpoint inhibitor response across melanoma patients used paired scRNA-seq and T-cell receptor sequencing to distinguish tumor-reactive T-cell clones from immunological bystanders. The technology has moved from proof-of-concept to routine deployment in cohort-scale studies.

Figure 2: Sample Preparation Decision Tree — Fresh Cells vs Nuclei vs Fixed Cells vs FFPE Workflow Selection

scRNA-Seq Core Technology

scRNA-seq workflows converge on a common architecture. A single-cell suspension is loaded onto a microfluidics instrument where individual cells are co-encapsulated with barcoded gel beads in oil droplets. Each bead carries millions of oligonucleotides sharing the same cell barcode but differing in a random unique molecular identifier (UMI), enabling transcript counting with molecular precision. Within each droplet, cells are lysed, mRNA is captured by poly(dT) priming, and reverse transcription incorporates the cell barcode and UMI into the cDNA. After droplet breaking and pooling, the barcoded cDNA from thousands of cells is amplified, fragmented, and sequenced in a single library.

The dominant commercial platform is the 10x Genomics Chromium series (1, 5). The current-generation GEM-X v4 chemistry improves transcript capture sensitivity compared to its predecessor, particularly for low-RNA cell types such as quiescent immune cells and certain neuronal populations. Cell capture efficiency runs approximately 50 to 65 percent, with doublet rates below 0.9 percent per 1,000 recovered cells. A standard run targeting 5,000 to 10,000 cells with 20,000 to 50,000 read pairs per cell provides robust clustering and differential expression for major cell populations present at frequencies above roughly one percent.

Several alternatives occupy important niches. SMART-seq3, a plate-based method, produces full-length transcript coverage rather than 3'- or 5'-biased reads, capturing splice junctions, isoform diversity, and point mutations. It excels when the number of cells is limited — hundreds rather than thousands — and deep characterization per cell matters more than population-level statistics. Parse Biosciences' split-pool barcoding approach eliminates microfluidics entirely, using sequential rounds of in-well barcoding to label up to 100,000 cells per experiment at a lower per-cell cost.

The emerging long-read frontier adds another dimension. PacBio HiFi and Oxford Nanopore sequencing now deliver full-length isoform resolution at single-cell resolution, resolving alternative splicing events, fusion transcripts, and allele-specific expression. This capability is particularly relevant for neuroscience, where alternative splicing generates extensive transcriptomic diversity, and for oncology, where gene fusions drive tumorigenesis.

For discovery projects where unbiased cell-type identification is the goal, 10x Genomics-based droplet scRNA-seq with 3' or 5' gene expression chemistry captures the full transcriptional landscape across thousands of cells in a single run. When the question targets an already-purified population — a sorted immune subset, a cultured line with a specific perturbation — bulk RNA-Seq often delivers deeper transcriptomic coverage at lower per-sample cost, and the marginal value of single-cell resolution diminishes. At the far end of the resolution spectrum, studies of alternative splicing, fusion transcripts, or allele-specific isoform usage increasingly combine single-cell capture with long-read sequencing on PacBio HiFi platforms, achieving full-length isoform resolution that short-read 3'-biased methods cannot provide.

For a detailed walkthrough of sample preparation and the scRNA-seq versus snRNA-seq decision, see our companion guide on Single-Cell and Single-Nucleus RNA Sequencing: Sample Preparation, Workflow, and Experimental Design for Frozen and Fresh Tissues.

Figure 3: 10x Genomics scRNA-seq Workflow — From Tissue Dissociation to Data Analysis

Spatial Transcriptomics

scRNA-seq's fundamental limitation is spatial blindness. Tissue dissociation erases all information about where each cell was located — whether a fibroblast sat adjacent to a tumor nest or in the distal stroma, whether two interacting immune cells were in physical contact, or whether a rare population clustered in a specific anatomical subregion. Spatial transcriptomics addresses this by capturing and sequencing mRNA directly from tissue sections, preserving the two-dimensional coordinates of each transcript (4, 7).

The technology landscape spans a resolution continuum. At the lower-resolution end, 10x Visium captures transcripts on slides printed with 55-micrometer spots, each containing spatially barcoded capture oligonucleotides. Each spot contains transcripts from approximately 1 to 10 cells. The Visium Fresh Frozen workflow uses poly(A) capture and works with any species — zebrafish, axolotl, plant species, and agricultural animals — making it the platform of choice for non-model organism research.

At the high-resolution end, Visium HD shrinks capture areas to 2-micrometer bins through a continuous oligo array, approaching single-cell resolution. The probe-based chemistry targets approximately 18,000 human or 20,000 mouse genes, making it suitable for FFPE cohorts where RNA is degraded and poly(A) capture would fail. Xenium takes a fundamentally different approach using single-molecule FISH with cyclic imaging to detect individual transcripts at subcellular resolution — approximately 0.2 micrometers per transcript — with panels covering up to 5,000 genes. Stereo-seq occupies a unique niche: 220-nanometer DNA nanoball-patterned arrays with whole-transcriptome coverage across capture areas up to 13 by 13 centimeters, large enough for whole mouse embryo sagittal sections (8).

The practical challenge is not understanding what each platform does — it is choosing among them. Most projects benefit from a layered strategy. Use Visium FF or Stereo-seq for discovery-level whole-transcriptome mapping, then validate with Visium HD for single-cell-scale resolution or Xenium for subcellular localization.

The layered strategy depends on unbiased access to multiple instruments. Researchers working with a provider that operates Visium Fresh Frozen, Visium HD, Xenium, and Stereo-seq on-site can design experiments around the biological question rather than around available hardware. This becomes decisive for projects spanning both discovery and validation: a whole-transcriptome survey on Visium FF identifies regions of interest, Visium HD delivers single-cell-scale resolution within those regions, and Xenium confirms key transcripts at subcellular level — all from serial sections of the same tissue block.

For an in-depth treatment of spatial transcriptomics method selection and FFPE compatibility, see our companion guide on 10x Genomics Spatial Transcriptomics for FFPE and Frozen Tissues: Mapping Gene Expression in Tissue Context.

Figure 4: Spatial Transcriptomics Platform Comparison — Visium vs. Visium HD vs. Xenium vs. Stereo-seq

Single-Cell Multi-Omics

Single-cell gene expression captures one dimension of cellular identity. But a T cell's functional state is defined not only by which genes it transcribes, but by the antigen receptor it expresses, the chromatin landscape that determines which genes are accessible, and the surface proteins that mark its lineage and activation status. Single-cell multi-omics integrates two or more of these data types from the same cell (6).

The 10x Genomics Single-Cell Multiome ATAC plus Gene Expression assay simultaneously measures chromatin accessibility and the transcriptome from the same nucleus. Isolated nuclei are transposed with Tn5 transposase, which preferentially inserts sequencing adapters into open chromatin regions, while nuclear mRNA is captured using barcoded gel bead chemistry. After sequencing, each cell barcode yields two linked data types — a gene expression profile and a chromatin accessibility profile. This paired readout enables gene regulatory network inference impossible with either modality alone.

Immune repertoire profiling adds another layer. The 5' gene expression assay captures the V(D)J region of T-cell receptor and B-cell receptor transcripts alongside the full transcriptome, enabling paired analysis of clonotype and cell state. A single T cell's transcriptome reveals whether it is activated, exhausted, or regulatory; its TCR sequence identifies which clones expanded in response to antigen.

CITE-seq (Cellular Indexing of Transcriptomes and Epitopes by sequencing) extends single-cell resolution to the protein level. Cells are stained with a panel of oligonucleotide-conjugated antibodies before encapsulation; each antibody carries a unique barcode that is captured alongside the cell's mRNA. A typical immune-profiling panel includes 30 to 50 antibodies targeting lineage markers (CD3, CD4, CD8, CD19, CD14), activation markers (CD25, CD69, HLA-DR), and checkpoint molecules (PD-1, TIM-3, LAG-3). Panel design must balance biological coverage against sequencing budget — each additional antibody consumes reads. For a 50-antibody panel on 10,000 cells, allocate approximately 5,000 to 10,000 additional reads per cell for antibody-derived tags on top of the transcriptomic read budget. Antibody titration is mandatory: titrate each antibody-conjugate individually to determine the staining index that cleanly separates positive from negative populations.

The practical bottleneck in multi-omics projects is rarely data generation — it is integration. Pairing chromatin accessibility with gene expression from the same nucleus, or linking a T cell's clonotype to its transcriptional phenotype, requires computational frameworks that go beyond standard scRNA-seq analysis pipelines. Weighted nearest neighbor analysis in Seurat v5 handles two modalities, but adding a third — protein abundance from CITE-seq antibody-derived tags, for instance — introduces dimensionality and batch-correction challenges that demand careful experimental planning upfront. Designing the computational integration strategy alongside the experimental workflow avoids the common scenario where multi-modal data accumulates faster than it can be meaningfully analyzed.

For a deeper treatment of immune repertoire profiling and CITE-seq experimental design, see our companion guide on Single-Cell V(D)J and CITE-Seq: Integrating Immune Repertoire Profiling and Multi-Omics at Single-Cell Resolution.

Figure 5: Single-Cell Multi-Omics Technologies — scRNA-seq, ATAC-seq, V(D)J, and CITE-Seq Integration

Experimental Design Decisions

The most consequential decisions in a single-cell project are made before the first cell enters the microfluidics chip. Cell number, sequencing depth, biological replicates, and platform choice are four interdependent variables that determine statistical power, biological resolution, and budget.

How Many Cells?

The answer is dictated by your rarest population of interest. If your target population represents roughly ten percent of input cells, capturing 3,000 total cells recovers approximately 300 target cells — adequate for differential expression. If your target is at one percent frequency, you need roughly 10,000 to 20,000 cells to recover 100 to 200 target cells, the minimum for reliable statistical comparison. For populations below 0.1 percent without enrichment, you need 50,000 to 100,000 or more captured cells.

A systematic evaluation across 23 datasets totaling 3.68 million cells established that approximately 500 cells per cell type per biological replicate is the minimum for reliable quantitative measurement of gene expression. Below this threshold, the coefficient of variation in pseudobulk expression estimates becomes unacceptably high. If you cannot reach 500 cells for your cell type of interest at a given budget, consider enrichment via FACS before library preparation.

How Many Reads Per Cell?

For cell-type classification, 20,000 to 25,000 read pairs per cell is sufficient. The marker genes that define lineages are strongly expressed, and additional sequencing shows diminishing returns for clustering resolution. For differential expression within a cell type, 50,000 read pairs per cell provides the sensitivity needed for moderate fold changes. For rare transcript detection — transcription factors, cytokines, non-coding RNAs — 50,000 to 100,000 or more reads per cell may be necessary.

The critical relationship is the cell-number-versus-depth trade-off. Within a fixed sequencing budget — say, 500 million read pairs — you can sequence 20,000 cells at 25,000 reads per cell, or 5,000 cells at 100,000 reads per cell. For cell atlas projects where discovering populations is the goal, allocate budget to cell number. For mechanistic studies where populations are already identified and deep transcriptional characterization is needed, allocate budget to depth per cell. Power analysis frameworks such as scPower model the interplay of sample size, cells per sample, and sequencing depth to guide these decisions (2).

Figure 6: Experimental Design Decision Framework — Cell Number vs. Sequencing Depth vs. Replicates Trade-off Matrix

Biological Replicates — The Non-Negotiable

The single most important statistical lesson from the first decade of scRNA-seq is that individual cells cannot be treated as independent replicates. Differential expression tests that compare individual cells between conditions treat each cell as an independent observation. But cells from the same biological sample share genetic background, environmental history, and technical processing batch. Treating them as independent inflates the effective sample size and produces false-positive rates of 30 to 80 percent (3).

The solution is pseudobulking. For each cell type within each biological replicate, aggregate the transcript counts across all cells to produce a single pseudobulk expression profile. Then apply standard bulk differential expression tools such as DESeq2, edgeR, or limma-voom. This approach models variation at the correct level — between biological replicates, not between individual cells — and reduces false-positive rates to the nominal 5 percent. Journals and reviewers increasingly require pseudobulk-based analysis for differential expression claims.

How many biological replicates? For a typical case-control comparison with moderate effect size — a two-fold change in 10 percent of genes within a cell type — three to five biological replicates per group, each with 500 or more cells of the target cell type, provides adequate power. Pilot experiments are strongly recommended: profile two to three samples first to determine cell-type frequencies and confirm the sample preparation workflow before committing the full cohort.

For a detailed walkthrough of cell number estimation, replicate planning, and budget optimization, see our companion guide on Experimental Design for Single-Cell Studies: Cell Number, Replicates, Sequencing Depth, and Cost Planning.

CD Genomics Service Workflow

A single-cell or spatial biology project spans sample preparation, library construction, sequencing, and data analysis. The CD Genomics workflow captures the key decision points where technical consultation adds the most value.

Figure 7: CD Genomics Single-Cell & Spatial Biology Service Workflow — Sample to Report

The process begins with a project consultation addressing the design questions above — cell number, depth, replicates, and platform selection. Sample preparation is the most technique-sensitive phase. Fresh tissue must be processed into a viable single-cell suspension within a narrow window after collection; enzymatic dissociation protocols are tissue-specific and must balance cell yield against transcriptional stress responses. For samples where intact cells cannot be obtained — frozen tissue, lipid-rich brain specimens, archived FFPE blocks — nuclei-based workflows bypass cell dissociation entirely, capturing nuclear transcripts from frozen and archived specimens that would be incompatible with whole-cell scRNA-seq.

Library preparation follows validated protocols with internal QC checkpoints. Sequencing is performed on Illumina NovaSeq platforms, with read configuration and depth matched to assay type and project goals. Bioinformatics analysis proceeds through Cell Ranger, Seurat v5, and ScanPy — covering demultiplexing, alignment, cell barcode filtering, ambient RNA removal, normalization, dimensionality reduction, clustering, and cell-type annotation. Downstream deliverables include interactive HTML reports with UMAP visualizations, cluster marker gene tables, and differential expression results.

For spatial transcriptomics, the analysis pipeline adds spatial-aware clustering, histology-guided spot selection, and integration with matched scRNA-seq data. For multi-omics projects, weighted nearest neighbor analysis in Seurat v5 integrates modalities within a shared dimensionality reduction. The computational demands are nontrivial: a standard 50,000-cell dataset requires 32 to 64 GB of RAM for preprocessing alone, and memory requirements scale nonlinearly with cell count. Dedicated bioinformatics infrastructure with Cell Ranger, Seurat v5, and ScanPy pipelines preconfigured handles this scaling automatically, letting researchers focus on biological interpretation rather than server configuration.

FAQ

When should I choose scRNA-seq over bulk RNA-seq?

Choose scRNA-seq when you need to resolve cellular heterogeneity — identifying cell types, detecting rare subpopulations, or tracing developmental trajectories. Choose bulk RNA-seq when you have purified a specific cell population and need deep transcriptomic profiling at lower cost per sample. For many-condition comparisons, consider "screen then zoom": bulk screening first, scRNA-seq on top hits.

What is the difference between scRNA-seq and snRNA-seq?

scRNA-seq uses intact cells from fresh tissue and captures cytoplasmic mRNA. snRNA-seq uses isolated nuclei from fresh, frozen, or difficult-to-dissociate tissue and captures nuclear transcripts, avoiding the transcriptional stress artifacts induced by enzymatic dissociation. snRNA-seq is the method of choice for frozen archived tissue and experiments requiring ATAC-seq from the same sample.

Which spatial transcriptomics platform should I choose?

Use Visium Fresh Frozen for whole-transcriptome discovery in any species, including non-model organisms. Use Visium HD for single-cell-resolution spatial mapping of human or mouse FFPE samples. Use Xenium for subcellular transcript localization with cell morphology readout. Use Stereo-seq for large-area mapping — whole embryos, entire organ cross-sections. Many projects benefit from combining two platforms: one for discovery, one for validation.

How many cells do I need to profile?

Target 500 cells per cell type per biological replicate as a minimum for quantitative analysis. Work backward from your rarest population of interest: if it represents 1 percent of input cells, capture at least 50,000 cells. If enrichment via FACS is feasible, the required capture number drops dramatically.

How many sequencing reads per cell do I need?

For cell-type classification, 20,000 to 25,000 read pairs per cell. For differential expression within a cell type, 50,000 read pairs per cell. For detecting low-abundance transcripts, 50,000 to 100,000 or more per cell. When in doubt, sequence a pilot sample and check the saturation curve.

Do I need biological replicates for single-cell experiments?

Yes. Individual cells cannot substitute for biological replicates. Use a minimum of three biological replicates per condition. Apply pseudobulking — aggregating counts within each cell type and replicate — then use DESeq2 or edgeR for differential expression testing. Analysis at the individual-cell level without pseudobulking produces false-positive rates of 30 to 80 percent.

Can CD Genomics handle non-model organisms?

Yes. The Visium Fresh Frozen and Stereo-seq platforms use poly(A) capture chemistry, which works with any species possessing a polyadenylated transcriptome — no species-specific probe panels required. We have processed samples from zebrafish, axolotl, rice, poplar, pig, cow, and numerous other non-standard research species.

What is the turnaround time for a typical single-cell project?

For a standard project of 4 to 8 samples, expect 3 to 4 weeks from sample receipt to sequencing completion, plus 2 to 4 weeks for full bioinformatic analysis. Expedited timelines are available.

References:

1. Luecken MD, Theis FJ. Current best practices in single-cell RNA-seq analysis: a tutorial. Molecular Systems Biology. 2019;15(6):e8746. doi:10.15252/msb.20188746
2. Schmid KT, Cruceanu C, Boettger A, et al. scPower accelerates and optimizes the design of multi-sample single cell transcriptomic studies. Nature Communications. 2021;12:6627. doi:10.1038/s41467-021-26779-7
3. Squair JW, Gautier M, Kathe C, et al. Confronting false discoveries in single-cell differential expression. Nature Communications. 2021;12:5692. doi:10.1038/s41467-021-25960-2
4. Du J, Yang YC, An ZJ, et al. Advances in spatial transcriptomics and related data analysis strategies. Journal of Translational Medicine. 2023;21:330. doi:10.1186/s12967-023-04150-2
5. Zheng GXY, Terry JM, Belgrader P, et al. Massively parallel digital transcriptional profiling of single cells. Nature Communications. 2017;8:14049. doi:10.1038/ncomms14049
6. Argelaguet R, Arnol D, Bredikhin D, et al. MOFA+: a statistical framework for comprehensive integration of multi-modal single-cell data. Genome Biology. 2020;21:111. doi:10.1186/s13059-020-02015-1
7. Fang S, Chen B, Zhang Y, et al. Computational approaches and challenges in spatial transcriptomics. Genomics, Proteomics & Bioinformatics. 2023;21(1):13-23. doi:10.1016/j.gpb.2022.09.006
8. Yue L, Liu F, Hu J, et al. A guidebook of spatial transcriptomic technologies, data resources and analysis approaches. Computational and Structural Biotechnology Journal. 2023;21:940-955. doi:10.1016/j.csbj.2023.01.016

For research purposes only, not intended for clinical diagnosis, treatment, or individual health assessments.