Written and maintained by CASRAI Editorial Board
Last updated
MSigDB, the Molecular Signatures Database, is a collection of annotated gene sets maintained by the GSEA-MSigDB team for use with gene set enrichment analysis (GSEA) and related methods. Its current release, v2026.2 (updated October 2026), exists as a Human MSigDB (v2026.2.Hs) and a Mouse MSigDB (v2026.2.Mm). Access to the gene sets requires a free registration. This guide lays out the collections with their current gene-set counts, the human and mouse mappings, how to cite the resource, and how it relates to tools such as Enrichr and clusterProfiler.
What MSigDB is
MSigDB is the gene-set library used by the GSEA software. Where GSEA is the analysis method, MSigDB supplies the gene sets that the method tests against a ranked expression list. The official site states that registration is free and is used to track usage for funding reports; registration is needed to download the GSEA software and the MSigDB gene sets and to use the web tools. Source (read 2026-11-07): gsea-msigdb.org MSigDB home.
Collections: Hallmark (H) and C1 to C9
The Human MSigDB is divided into the Hallmark collection and numbered collections. The counts below are for release v2026.2.Hs, as listed on the official collections page (read 2026-11-07). The whole Human MSigDB holds 35,092 gene sets.
| Collection | Name | Gene sets (v2026.2.Hs) | Subcollections listed |
|---|---|---|---|
| H | Hallmark | 50 | None |
| C1 | Positional | 302 | None |
| C2 | Curated | 7,837 | CGP 3,560; CP 4,277 (BioCarta 292, KEGG_MEDICUS 790, PID 196, Reactome 1,867, WikiPathways 927, KEGG_LEGACY 186) |
| C3 | Regulatory target | 3,712 | MIR 2,598 (MIRDB 2,377, MIR_LEGACY 221); TFT 1,114 (GTRD 504, TFT_LEGACY 610) |
| C4 | Computational | 1,006 | 3CA 148; CGN 427; CM 431 |
| C5 | Ontology | 15,849 | GO 9,917 (BP 6,905, CC 1,094, MF 1,918); HPO 5,932 |
| C6 | Oncogenic signature | 189 | None |
| C7 | Immunologic signature | 5,219 | ImmuneSigDB 4,872; VAX 347 |
| C8 | Cell type signature | 866 | None |
| C9 | Computational perturbation (outside H and C1-C8) | 62 | DepMap CRISPR and CCLE |
Older primers often stop at C7 or C8; C9 appears on the current collections page. Source (read 2026-11-07): MSigDB collections page.
MSigDB Hallmark gene sets
The Hallmark collection (H) has 50 gene sets. The official description is that each represents a well-defined biological state or process with coherent expression, built by overlapping gene sets from other collections. That makes H a compact collection to start from, compared with the 35,092 sets in the full database.
MSigDB gene sets: choosing a subset
The collections page advises against running an analysis on the full Human MSigDB and recommends using focused collections. In practice that means choosing one collection or subcollection whose description matches your question: for example, C5 for Gene Ontology and Human Phenotype Ontology terms, C7 for immunologic signatures, or C2 for curated pathway resources such as Reactome and WikiPathways. For how GSEA consumes a gene set file, see the GMT section of the GSEA guide.
Human and mouse MSigDB
MSigDB is published in two parts, Human MSigDB (v2026.2.Hs) and Mouse MSigDB (v2026.2.Mm). The mouse collections carry their own labels: MH (mouse Hallmark), M1, M2, M3, M5, M7 and M8. The official page lists no mouse equivalent for C4 or C6. Source (read 2026-11-07): MSigDB home.
MSigDB download
Gene sets are downloaded from the MSigDB site after free registration, and the collections page shows what each collection contains before you download. The sign-in is linked from the MSigDB home page. Gene sets are distributed for GSEA as GMT files, described in the GSEA guide.
MSigDB in R: the msigdbr package
msigdbr is a third-party R package, not run by the GSEA-MSigDB team, that provides MSigDB gene sets as a tidy data frame. It supplies the human genes as in MSigDB plus matching symbols and identifiers for mouse, rat, pig, fly and yeast. Version 26.1.1 was published on CRAN on 2026-08-20; it requires R 4.1 or later, is MIT-licensed, and is maintained by Igor Dolgalev. Source (read 2026-11-07): msigdbr on CRAN. Gene sets obtained this way can be analysed with enrichment tools such as clusterProfiler.
MSigDB citation
The MSigDB home page asks users to cite:
- Subramanian et al. 2005, PNAS (the GSEA method paper);
- one or more of Liberzon et al. 2011, Bioinformatics, and Liberzon et al. 2015, Cell Systems (the MSigDB papers);
- the source listed on each gene set page; and
- for Mouse MSigDB, also Castanza et al. 2023, Nature Methods.
Also state the release (for example v2026.2.Hs) and the collection used, since gene-set contents differ between releases. Source (read 2026-11-07): MSigDB home.
Strengths and limits
- Strength: a single, versioned, citable source of gene sets spanning curated pathways, ontologies, regulatory targets, and cancer, immune and cell-type signatures, with a separate mouse resource.
- Strength: Hallmark (50 sets) gives a small, coherent starting collection.
- Limit: the full database (35,092 human sets) is not meant to be tested in one run, according to the official collections page.
- Limit: registration is required to download gene sets and use the web tools.
Alternatives and companions
| Resource | What it is | Relation to MSigDB |
|---|---|---|
| Enrichr | Web-based enrichment analysis tool | See the Enrichr guide for how it differs from running GSEA |
| clusterProfiler | R package for enrichment analysis | Can be paired with gene sets obtained through msigdbr |
| Reactome | Pathway knowledgebase | Source of 1,867 of the C2 canonical pathway sets in v2026.2.Hs |
| KEGG | Pathway resource | Represented in C2 as KEGG_MEDICUS (790) and KEGG_LEGACY (186) |
FAQ
What does MSigDB stand for?
Molecular Signatures Database, the gene-set resource maintained by the GSEA-MSigDB team.
How many gene sets are in MSigDB?
The Human MSigDB release v2026.2.Hs lists 35,092 gene sets across the Hallmark collection and C1 to C9, per the official collections page read on 2026-11-07.
Is MSigDB free?
Registration is free. The site says registration is used to track usage for funding reports.
Is there an MSigDB for mouse?
Yes. Mouse MSigDB v2026.2.Mm uses the labels MH, M1, M2, M3, M5, M7 and M8.
Which MSigDB paper should I cite?
Cite Subramanian 2005 (PNAS), at least one of Liberzon 2011 or Liberzon 2015, and the source on each gene set page; add Castanza 2023 for mouse.
Sources
- MSigDB home, gsea-msigdb.org (read 2026-11-07)
- MSigDB collections, gsea-msigdb.org (read 2026-11-07)
- msigdbr, CRAN (read 2026-11-07)
