Written and maintained by CASRAI Editorial Board
Published
On this page
The COSMIC database, short for the Catalogue of Somatic Mutations in Cancer, is a curated resource run by the Wellcome Sanger Institute that gathers somatic mutations found in human cancers, with the literature and tumour-sample context behind each record. It is the standard reference for asking which genetic changes have been reported in which cancers, and it is not related to cosmic physics or astronomy.
What COSMIC is
COSMIC describes itself as “the world’s most comprehensive resource for exploring the impact of somatic mutations in human cancer” (COSMIC About page, read 2026-11-20). A somatic mutation is a change that arises in body cells during life and is found in a tumour but not inherited, which is why a catalogue of them is useful to cancer genomics researchers, variant curators and anyone building a pipeline that annotates tumour sequencing data.
The resource was founded in 2004 in the Cancer Programme at the Wellcome Sanger Institute, associated with the investigators Michael Stratton and Andrew Futreal. Its core method, described in the 2019 Nucleic Acids Research update by Tate and colleagues, is expert manual curation: trained curators extract mutation data from the published literature, and the team also imports data from large systematic screens. New releases appear every three months.
COSMIC is better thought of as a family of linked projects than a single table. The download page names the Cancer Gene Census, Genome Screen Mutants, Breakpoints, Cell Lines, COSMIC 3D, the Cancer Mutation Census and Actionability as separate downloadable products, and the site menu groups browsing under a Cancer Browser and a Genome Browser. The Cell Lines Project, which characterises more than 1,000 cancer cell lines, is described on the Sanger Institute’s COSMIC tool page.
Who runs it and what it covers
COSMIC is operated by the Wellcome Sanger Institute; see the CASRAI guide to the Wellcome Sanger Institute for the institute’s wider programmes and data portals. The Institute states that COSMIC is now under licence, with academic access available freely and commercial access available for a licence fee, and that licence revenue funds curation. The COSMIC site names QIAGEN as the commercial partner for commercial users.
The most recent peer-reviewed description of the database’s content is Tate et al. (2019), which reports on release v86 of August 2018. The figures below describe that release only. Later releases have grown the catalogue, so treat these as a dated snapshot of scope, not as current totals. The data table below the article compiles them.
In v86, the coding-mutation core held about 5.98 million coding mutations from about 1.39 million tumour samples, curated from about 26,251 publications. Beyond coding point mutations, the same release covered 19.7 million non-coding variants, 1.18 million copy-number variants, 19,368 gene fusions, 9.15 million differential expression variants and 7.88 million differentially methylated CpG sites. It also included data from 35,480 whole genomes and exomes drawn from 457 studies, a drug-resistance dataset covering 24 drugs, 360 alleles and 2,134 resistant tumours, and a Cell Lines Project with 1,015 lines.
COSMIC-3D
Tate et al. describe COSMIC-3D as a view that maps missense mutations, in-frame deletions and nonsense mutations onto three-dimensional protein structures, including drug-binding pockets, so a variant can be inspected in the context of the protein it changes. It is also listed as a separate download product.
Cancer Gene Census
The Cancer Gene Census (CGC) is the curated list of genes for which there is evidence of a causal role in cancer. In v86 it contained 719 genes. Genes are split into two tiers. Tier 1 genes have strong evidence of involvement in cancer, supported by independent research groups. Tier 2 genes have extensive literature support that is less robust. The Census annotates genes to the hallmarks of cancer, which lets a researcher filter a gene list by the biological capability a gene contributes to.
The Census gene count is also a release-specific number. On the download page, the Cancer Gene Census can be filtered by gene symbol only, whereas other products allow more filters, so plan to download it whole when you need the full list.
Downloading data and licences
COSMIC data are not freely redistributable to everyone. The download page states that Non-Commercial and Commercial licences are offered, that commercial users should contact QIAGEN, and that non-commercial users should register and log in to download. The sign-in and registration links are on the COSMIC site (cancer.sanger.ac.uk/cosmic); this guide does not cover account steps beyond pointing to them.
| Question | Non-commercial licence | Commercial licence |
|---|---|---|
| Who it is for | Academic and other non-commercial users | Commercial users |
| Cost | Academic access is available freely | A licence fee applies |
| How to get access | Register, then log in to download | Contact QIAGEN |
| Licensing enquiries (Sanger) | [email protected], per the Sanger COSMIC tool page | |
Three ways to download
The download page, with the genome build selectable (GRCh37 shown, GRCh38 the alternative), describes three modes:
- Download in Browser delivers a tar archive of zipped tab-separated files plus a readme.
- Scripted Downloads is meant for automated retrieval, for example in a pipeline.
- Filtered Downloads lets you restrict a file by gene symbol, sample name or primary site; it is available for the latest release only.
Because filtered downloads cover the latest release only, record the release version in your methods when you use them. A study that cites a COSMIC annotation without a release number cannot be reproduced exactly, since the catalogue is revised every three months. For reproducibility practice in general, see the CASRAI guide to maftools, which analyses mutation annotation format files in R and is a common downstream consumer of somatic-mutation data.
Browsing the catalogue
Under the Tools menu, the COSMIC site offers a Cancer Browser and a Genome Browser, so records can be explored by cancer type or in genomic context. A Search item sits in the main navigation. The site also lets you choose the genome version, GRCh38 or GRCh37, so confirm which build your own coordinates use before comparing positions. Step-by-step search syntax is documented in the COSMIC Help section, which is the place to check for current interface wording.
A worked example: documenting a COSMIC download
Suppose a research group annotates tumour sequencing variants with COSMIC data. A reproducible methods note records the genome build (GRCh37 or GRCh38, the two options on the download page), the product used (for example the Cancer Gene Census or Genome Screen Mutants), the download mode, the release version and the date of download, and the licence category under which the data were obtained. If a filtered download was used, the note should also state the filter (gene symbol, sample name or primary site). This is an editorial suggestion built on the facts above, not a COSMIC requirement, but each element it lists is something the catalogue’s three-monthly changes make necessary to pin down.
Who tends to use which part
- Cancer genomics researchers exploring by cancer type or genome position can use the Cancer Browser and the Genome Browser.
- Structural biologists and drug-discovery teams can use COSMIC-3D to place variants on protein structures and drug-binding pockets.
- Groups building gene panels or filters often start from the Cancer Gene Census tiers.
- Commercial teams need the commercial licence route before using the data in a product or service.
Strengths and limits
Strengths
- Curated from the literature. Expert curators extract records from published articles, and large screens are imported in addition, so each record can be traced to a study.
- Several data types in one place. Coding mutations, non-coding variants, copy-number variants, fusions, expression and methylation are covered in the v86 description.
- A curated gene list. The Cancer Gene Census gives a tiered, evidence-graded list of cancer genes that many analyses use as a reference set.
- Regular releases. The three-monthly cycle means new literature is folded in on a predictable schedule.
Limits
- Licensing. Access is free for academic use only; commercial use needs a paid licence, and reuse of downloaded data is governed by the licence that applies to you.
- Release drift. Counts and gene numbers change between releases, so a figure quoted from the 2019 paper does not describe today’s catalogue.
- Not a clinical classification. COSMIC records reported somatic mutations. It does not present itself as grading variants for clinical actionability in the way a knowledge base such as OncoKB does, although an Actionability product is listed among the downloads.
- Filtering is narrow for some products. The Cancer Gene Census filter is by gene symbol only.
Alternatives and related resources
These are other resources that CASRAI has covered. They answer different questions from COSMIC, so they are complements more often than replacements.
| Resource | What it is, per the CASRAI guide | Guide |
|---|---|---|
| OncoKB | Memorial Sloan Kettering’s precision oncology knowledge base, with evidence levels, an Annotator and an API | OncoKB guide |
| HGMD | The Human Gene Mutation Database, with Public and Professional versions and defined variant classes | HGMD guide |
| ClinGen Allele Registry | Canonical allele identifiers (CAids) with sequence normalisation and an API | ClinGen Allele Registry guide |
| maftools | An R package for analysing mutation annotation format files and drawing oncoplots | maftools guide |
Frequently asked questions
What does COSMIC stand for?
COSMIC stands for the Catalogue of Somatic Mutations in Cancer. It is run by the Wellcome Sanger Institute and is unrelated to astronomy, despite the shared word.
Is the COSMIC database free?
It is free for academic and other non-commercial users, who register and log in to download. Commercial users need a licence for a fee, arranged through QIAGEN, according to the COSMIC download page and the Sanger Institute’s COSMIC tool page.
How often is COSMIC updated?
Releases appear every three months, according to the Sanger Institute and the 2019 Nucleic Acids Research update. Note the release number whenever you use data, because filtered downloads cover the latest release only.
What is the Cancer Gene Census?
It is COSMIC’s curated list of genes with evidence of a causal role in cancer, split into Tier 1 (strong evidence from independent groups) and Tier 2 (extensive but less robust literature). It held 719 genes in release v86.
Where does the data come from?
Expert curators extract mutations from published literature, and the team also imports data from large systematic screens, as described by Tate et al. (2019).
Which file formats can I download?
The browser download is a tar archive of zipped tab-separated files with a readme. Scripted and filtered downloads are the other two modes on the download page.
Sources
- COSMIC: About (read 2026-11-20)
- COSMIC: Download page (read 2026-11-20)
- Wellcome Sanger Institute: COSMIC tool page (read 2026-11-20)
- Tate JG et al. COSMIC: the Catalogue Of Somatic Mutations In Cancer. Nucleic Acids Research 2019;47(D1):D941-D947 (read 2026-11-20)
Data: COSMIC release v86 (August 2018): record counts by data type
Counts of records by data type in COSMIC release v86, compiled from the Tate et al. 2019 Nucleic Acids Research update describing that release. Figures are release-specific and have grown in later releases.
Period covered: 2018.
| Data type | Count | Note |
|---|---|---|
| Coding mutations | 5980000 | Approximate; from about 1.39 million tumour samples |
| Tumour samples | 1390000 | Approximate |
| Curated publications | 26251 | Approximate |
| Non-coding variants | 19700000 | Approximate |
| Copy-number variants | 1180000 | Approximate |
| Gene fusions | 19368 | As reported |
| Differential expression variants | 9150000 | Approximate |
| Differentially methylated CpG sites | 7880000 | Approximate |
| Whole genomes and exomes | 35480 | From 457 studies |
| Cancer Gene Census genes | 719 | Tier 1 and Tier 2 combined |
| Cell Lines Project lines | 1015 | As reported |
Sources
- Tate JG et al. COSMIC: the Catalogue Of Somatic Mutations In Cancer. Nucleic Acids Res 2019 (accessed 2026-11-20)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







