Written and maintained by CASRAI Editorial Board
Last updated
Short answer. Single-cell RNA sequencing (scRNA-seq) measures gene expression in individual cells rather than as an average over a tissue. Cells are captured and their RNA is tagged with cell barcodes and unique molecular identifiers (UMIs). The libraries are sequenced, and software turns the reads into a cell-by-gene count matrix for quality control and analysis.
Definition and scope
scRNA-seq is used to characterize gene expression differences between individual cells derived from a complex tissue (PacBio, “Single-cell RNA sequencing”, read 2026-11-07). That separates it from bulk RNA-seq, where one measurement per sample hides differences between cell types. Wolfien, David and Galow describe the overall workflow in four phases: tissue and cell preparation; cell capture and library preparation; sequencing and raw data processing; and visualization and downstream analysis (Exon Publications, 2021, CC BY-NC 4.0, read 2026-11-07).
This page explains the workflow and gives two checklists. It does not cover cost, which is on a separate page: see single-cell sequencing cost.
How the scRNA-seq workflow works
Principle: barcodes and UMIs
Each cDNA molecule is tagged with a cell barcode and a UMI, so that after pooled sequencing every read can be traced back to its cell of origin (PacBio). Tag-based protocols combined with UMIs permit multiplexing and improve quantification (Wolfien et al.).
Steps, from sample to count matrix
- Tissue and cell preparation. Tissue is turned into a suspension of single cells or nuclei.
- Cell capture and library preparation. Cells are isolated in droplets, microwells or microfluidic chambers, and their RNA is barcoded and converted to a library.
- Sequencing and raw data processing. Reads are aligned and assigned to cell barcodes and UMIs to produce a count matrix. Cell Ranger is named as one tool that produces the count matrix, and Kallisto is lighter on compute than STAR at the cost of some information (Wolfien et al.).
- Visualization and downstream analysis. Normalization, clustering and visualization, then trajectory or pseudotime analysis, RNA velocity or network inference.
Analysis pipeline and workflow in R or Python
Wolfien et al. classify standard data processing into six stages: (i) raw data alignment; (ii) quality control and data normalization; (iii) data integration and correction; (iv) expression recovery; (v) feature selection; and (vi) dimensionality reduction and visualization. They call Seurat (R) one of the most popular toolboxes, and also name Scanpy (Python) and Scater (R). For batch integration they name Harmony, Liger and Seurat 3, and report Harmony as the first to try in a benchmark because of its shorter runtime. Follow-on topics have their own guides: Leiden clustering for cell grouping, differential gene expression analysis, and bioinformatics in Python.
Types: capture formats compared
The table is limited to what Wolfien et al. state about each approach, plus one attributed vendor statement.
| Approach | What the source states | Question it raises |
|---|---|---|
| Droplet-based (Drop-seq, inDrop, 10x Genomics Chromium) | Drop-seq and inDrop: about 5,000-10,000 cells, moderate sensitivity. 10x Chromium: 1,000-10,000 cells, high sensitivity. | Do you need thousands of cells per run? |
| Microwell or plate-based (Smart-seq2, CEL-seq2) | Fewer than 400 cells; Smart-seq2 is listed with very high sensitivity. | Do you need deep per-cell coverage from few cells? |
| Microwell or microfluidic with imaging | Microscope inspection can exclude empty wells, damaged cells or doublets; high cartridge cost can limit sample size. | Is visual confirmation of each capture worth a smaller sample? |
| Tag-based (one transcript end) with UMIs | Restriction to one end of the transcript may hamper alignment and makes these protocols unsuitable for allele-specific expression or isoform usage. | Do you need isoforms or alleles? |
| Long-read full-length (PacBio, vendor statement) | PacBio states HiFi reads can cover the whole transcript, giving isoform information per cell; short reads sequence a short fragment of one end, giving gene-level information. | Is isoform diversity per cell the question? |
For long-read background, see the CASRAI pages on PacBio sequencing and Oxford Nanopore.
Where it matters for labs: a QC checklist
Wolfien et al. name three QC covariates calculated per barcode. Interpret them together, and treat thresholds as study-specific.
| Covariate | Pattern | What it may indicate (per the source) |
|---|---|---|
| Counts per barcode (count depth) | Low | Low-quality cell; together with the two rows below it may flag leaked cytoplasmic mRNA |
| Genes per barcode | Few | Same pattern as low depth |
| Fraction of counts from mitochondrial genes | High | Broken membrane with cytoplasmic mRNA lost |
| Counts per barcode | Very high | May represent doublets; tools such as Scrublet or Solo can filter them |
Caution from the same source: heterogeneous populations differ in total mRNA content, so a single cutoff can remove genuine cell types. RNA quality before capture is a separate matter; see RNA integrity number. Public datasets are commonly deposited in repositories such as the Gene Expression Omnibus.
Common misunderstandings
- It is not bulk RNA-seq at smaller scale. The unit of measurement is the cell, with barcodes and UMIs to keep cells and molecules apart.
- Short-read data gives gene-level information. Isoform and allele-specific questions need full-length data and are unsuited to tag-based protocols (Wolfien et al.; PacBio).
- A count matrix is not clean data. It still needs QC, normalization and, across batches, integration and correction. Dropout imputation methods exist, for example scTSSR (Wolfien et al.).
FAQ
What is single-cell RNA sequencing used for?
To characterize gene expression differences between individual cells from a complex tissue (PacBio).
What are the main steps?
Tissue and cell preparation, cell capture and library preparation, sequencing and raw data processing, then visualization and downstream analysis (Wolfien et al.).
Which software is used for analysis?
Seurat, Scanpy and Scater are named, with Cell Ranger for count matrices (Wolfien et al.).
Sources
- Wolfien M, David R, Galow A-M. Single-Cell RNA Sequencing Procedures and Data Analysis. In: Bioinformatics. Exon Publications; 20 March 2021. exonpublications.com. Read 2026-11-07.
- PacBio. Single-cell RNA sequencing. pacb.com. Vendor page; read 2026-11-07.
