BIOINFORMATICS APPLICATIONS NOTE
Vol. 27 no. 12 2011, pages 1739–1740
Databases and ontologies
Molecular signatures database (MSigDB) 3.0
Arthur Liberzon, Aravind Subramanian, Reid Pinchback, Helga Thorvaldsdóttir,
Pablo Tamayo and Jill P. Mesirov∗
Broad Institute of MIT and Harvard, Cambridge, MA 02142, USA
Associate Editor: Alex Bateman
Advance Access publication May 5, 2011
Motivation: Well-annotated gene sets representing the universe of
the biological processes are critical for meaningful and insightful
interpretation of large-scale genomic data. The Molecular Signatures
Database (MSigDB) is one of the most widely used repositories of
Results: We report the availability of a new version of the database,
MSigDB 3.0, with over 6700 gene sets, a complete revision of the
collection of canonical pathways and experimental signatures from
publications, enhanced annotations and upgrades to the web site.
Availability and Implementation: MSigDB is freely available for
non-commercial use at http://www.broadinstitute.org/msigdb.
Received on January 13, 2011; revised on April 4, 2011; accepted
on April 12, 2011
Microarrays and other high-throughput genomic technologies
typically produce long lists of potentially interesting genes, which
are not always easily interpreted. Recognizing the importance of
coordinately expressed sets of genes, our seminal paper (Mootha
et al., 2003) introduced Gene Set Enrichment Analysis (GSEA)
to discover metabolic pathways altered in human type 2 diabetes
mellitus. GSEA and other analytical enrichment tools summarize
genomic data in prioritized lists of higher-level biological features.
As underscored by a recent survey of 68 enrichment tools, they
critically depend on ‘backend annotation databases’ (Huang et al.,
2009). Typically, such databases focus on a particular domain of
knowledge or annotation procedure. For example, Gene Ontology
(GO) (Ashburner et al., 2000) represents a hierarchy of controlled
terms to describe individual gene products, while TRANSFAC
(Matys et al., 2006) stores information about transcription factor
binding sites. A growing number of databases obtain sets from
gene expression signatures reported in the literature. These include
SignatureDB (Shaffer et al., 2006), GeneSigDB (Culhane et al.,
2009), CCancer (Dietmann et al., 2010) and L2Land LOLA(Cahan
et al., 2007).
Molecular Signatures Database (MSigDB) differs from these
resources in several distinguishing aspects. (i) MSigDB is explicitly
designed to provide gene sets for enrichment analysis methods.
As such, it is natively and seamlessly integrated with our GSEA
software (Subramanian et al., 2005). (ii) MSigDB covers a
substantially more diverse and wider range of gene set sources
and types. These include signatures extracted from original research
∗To whom correspondence should be addressed.
publications, and entire collections of sets derived from specialized
resources such as GO, KEGG (Kanehisa and Goto, 2000),
TRANSFAC and L2L. (iii) MSigDB gene sets are acquired both
through manual curation and by automatic computational means,
whereas other databases emphasize only one of these approaches.
(iv) Finally, MSigDB contains the largest number of gene sets
The initial MSigDB database, released in 2005 with GSEA
software, contained 1325 sets. In contrast, MSigDB 3.0, released in
September 2010, includes 6769 sets and a richer set of annotations.
Here, we describe the MSigDB 3.0 sets in more detail and the
accompanying online resource.
Gene set collections: gene sets in MSigDB 3.0 are organized into
five collections according to their derivation:
C1: Genes located in the same chromosome or cytogenetic band.
C2: Gene sets representing canonical pathways from pathway
resources [including 430 new sets contributed by Reactome
(Matthews et al., 2009)], and sets corresponding to chemical
and genetic perturbations from 786 scientific publications.
C3: Sets of genes sharing cis-regulatory motifs in their promoter
(transcription factor targets) or 3?UTR (micro-RNA targets)
C4: Clusters of coexpressed modules defined by computational
analysis of large gene expression compendia.
C5: Gene sets corresponding to GO terms.
since the initial release (see also online Release Notes).
Gene set annotations: each MSigDB gene set is a list of genes
with relevant annotations and links to external resources. MSigDB
focuses on human gene sets. However, we do include sets from
some model organisms and gene set annotations include organism
identification. We use HUGO gene symbols and, as of version 3.0,
human Entrez Gene IDs serve as universal identifiers. These Entrez
IDs are guaranteed to be unique and stable, can easily be mapped
into a variety of other identifiers and are natively integrated with
the GenBank resources of primary nucleic and protein sequences.
We also preserve whatever original identifiers were used in the gene
set source. All sets have unique database identifiers and names, and
include brief and full descriptions. Other annotations depend on
the type of gene set. Annotations linking to external resources are
especially important as they allow researchers to place the sets in
© The Author 2011. Published by Oxford University Press. All rights reserved. For Permissions, please email: firstname.lastname@example.org
A.Liberzon et al.
Table 1. MSigDB versions and changes in the number of gene sets
Gene set category 1.0 (2005)2.5 (2008) 3.0 (2010)
C2: curated (total)
C2: chemical and
C2: canonical pathways
C3: motifs (total)
C3: transcription factor
C3: micro-RNA targets
C5: GO terms
aDecrease in number due to the removal of sets with too few genes to run GSEA.
the context of a specific study and facilitate decisions on follow-up
Gene sets from publications are the most richly annotated. Their
annotations include the PubMed ID of the publication, pointers to
other gene sets from the same publication, and now also details on
the exact table or figure from which the gene set was extracted.
For version 3.0, we updated the names of these gene sets to make
them more descriptive and standardized and the accompanying brief
descriptions to follow a more uniform and consistent format. Other
annotation features introduced with version 3.0 include links to
source datasets in Gene Expression Omnibus (GEO) (Barrett et al.,
sets include links to the pathway at the source web site.
File formats: MSigDB gene set files are available for download
in plain text and XML formats. The plain text files contain simple
listings of gene set membership, while the XML files also include
the annotations. To ensure reproducibility of GSEA results, older
versions of the MSigDB files are always available. Note that users
of our GSEA software do not need to download the MSigDB files
as the tool directly and automatically retrieves the gene sets.
In version 3.0, we updated the MSigDB web site. First introduced
in July 2007, the site allows users to view the annotated gene
sets and perform simple search and analysis tasks. Each gene set
and all of its annotations are presented on a separate web page
external web resources, including PubMed, GEO andArrayExpress,
PubChem and Entrez Gene.
The MSigDB web site allows users to find gene sets by
searching for keywords in the annotations. The online analysis
tools allow users to: (i) compute overlaps between gene sets; (ii)
view a heat map of a gene set in one of the reference expression
compendia; and (iii) categorize the genes in a set by gene families.
Gene families offer a quick view of a gene set by grouping its
members into a small number of informative categories. We have
updated the gene families and they now include: oncogenes, tumor
suppressors, translocated cancer genes, transcription factors, protein
MSigDB ONLINE RESOURCE
Fig. 1. A typical gene set page on the MSigDB web site. The list of genes
has been abbreviated from 41 to 2 for the purposes of this figure.
kinases, homeodomain proteins, cell differentiation markers and
We thank J. Roberston, L. Saunders and L. Kazmierski for gene
set collection; H. Kuehn and J. McLaughlin for documentation; and
M. Wrobel for web site development.
Funding: National Cancer Institute (5R01CA121941).
Conflict of Interest: none declared.
Ashburner,M. et al. (2000) Gene ontology: tool for the unification of biology. The Gene
Ontology Consortium. Nat. Genet., 25, 25–29.
Barrett,T. et al. (2009) NCBI GEO: archive for high-throughput functional genomic
data. Nucleic Acids Res., 30, D5–D15.
Cahan,P. et al. (2007) Meta-analysis of microarray results: challenges, opportunities,
and recommendations for standardization. Gene, 401, 12–18.
Nucleic Acids Res., 38, D716–D725.
Dietmann,S. et al. (2010) CCancer: a bird’s eye view on gene lists reported in cancer-
related studies. Nucleic Acids Res., 38 (Suppl), W118–W123.
Huang,da W. et al. (2009) Nucleic Acids Res., 37, 1-13.
Kanehisa,M. and Goto,S. (2000) KEGG: Kyoto encyclopedia of genes and genomes.
Nucleic Acids Res., 28, 27–30.
Matthews,L. et al. (2009) Reactome knowledgebase of human biological pathways and
processes. Nucleic Acids Res., 37, D619–D622.
regulation in eukaryotes. Nucleic Acids Res. 34, D108–D110.
Mootha,V. et al. (2003) PGC-1alpha-responsive genes involved in oxidative
phosphorylation are coordinately downregulated in human diabetes. Nat. Genet.,
Parkinson,H. et al. (2009) ArrayExpress update – from an archive of functional
genomics experiments to the atlas of gene expression. Nucleic Acids Res., 37,
Shaffer,A. et al. (2006) A library of gene expression signatures to illuminate normal
and pathological lymphoid biology. Immunol Rev., 210, 67–85.
Subramanian,A. et al. (2005) Gene set enrichment analysis: a knowledge-based
approach for interpreting genome-wide expression profiles. Proc. Natl Acad. Sci.
USA, 102, 15545–15550.