Seurat

Seurat, developed in the Satija Lab at the New York Genome Center, is the R toolkit that defined how single-cell RNA-seq analysis is done. Its clustering and integration workflows became the field’s default vocabulary, and a very large share of published single-cell figures were produced with it.

What it does

The standard pipeline runs from raw counts through quality control, normalisation, variable feature selection, scaling, PCA, neighbour graph construction, clustering and UMAP or t-SNE visualisation, then on to differential expression and marker identification. Its integration methods — anchor-based correction and, more recently, the reference-mapping workflow — are the reason many labs chose it, allowing datasets from different batches, donors, technologies or species to be analysed together.

Seurat has extended well beyond transcriptomics into multimodal work: CITE-seq protein and RNA together, ATAC alongside expression, and spatial transcriptomics from the major commercial platforms.

Why it matters

Statistical genomics has a deep R tradition, and Seurat sits in that ecosystem alongside Bioconductor, limma, edgeR and DESeq2. For labs whose analysis culture is already R-based, Seurat means single-cell work happens in the same language, with the same plotting conventions, as everything else they do.

Strengths

• Free, open source, and extremely widely used, so protocols and troubleshooting are everywhere.
• Best-in-class integration and reference-mapping workflows.
• Strong multimodal support across CITE-seq, ATAC and spatial data.
• Native to the R statistical ecosystem.
• Vignettes are unusually good, and the surrounding literature is vast.
• ggplot2-based plotting integrates with existing figure pipelines.

Limitations

Memory handling is its weak point — large datasets can be punishing in R, and Scanpy generally handles the largest experiments more gracefully. Major version transitions have broken object structures and workflows more than once, which is painful for long-running projects. The convenience of the default pipeline invites uncritical use, and clustering parameters chosen because they were in the vignette are a recognisable problem in the published literature.

Verdict

The right choice for R-based labs and still the strongest option for integration-heavy work. Pin your Seurat version at the start of a project and record it, because the object model has changed before and will change again.