Background. COSMIC (http://www.sanger.ac.uk/cosmic) is a system designed to curate the world's literature on somatic mutations in known cancer genes. Initially conceived to capture the mutation spread in point-mutated genes, COSMIC has now grown to encompass gene fusion products of genome rearrangement events which generate completely novel transcripts, together with all the somatic mutation data from candidate gene screens at the Cancer Genome Project, UK (CGP), covering almost 5000 genes of potential interest in cancer genetics. Results. The latest release of COSMIC (version 37; July 2008) now holds full and up-to-date curation of over 5,900 scientific papers, examining over 268,000 tumours, in which over 59,000 mutations are detailed through 60 pointmutated genes. Fusion gene products have been curated for 16 pairs of genes, described through over 4200 tumours. 2246 papers were rejected during manual curation, usually due to significant inconsistencies in the publication. A relational database holds the captured information, which is warehoused for each release. The information is presented on the internet with a series of graphical and tabulated views aiding navigation and interpretation. Conclusions. The current version of COSMIC is close to fulfilling its original intentions, with curation of most pointmutated genes in cancer complete. However, new challenges are emerging with the need to calculate the effect of high numbers of observed sequence changes to identify those driving tumour formation, and the need to meaningfully handle the increasing quantities of data from high-throughput screens and next-generation sequencing technologies.
The scientific literature contains tens of thousands of publications describing the involvement of somatic gene mutations in a wide range of human cancer phenotypes and this is enhanced by a number of online resources. However, these online resources tend to be focused on individual loci (e.g. IARC p53 database, The wide distribution of the data in the literature, together with the broad spread of available details and formats makes it very difficult to investigate for aggregate statistical interpretation. COSMIC is designed to overcome this limitation by curating all this data, in as much detail as possible, into one repository which can be examined easily on the internet. Literature curation data are presented in tandem with somatic [1]), not providing genome-wide information on mutation combinations within tumours, or very broadly focused, storing minimal information, usually only on high-frequency mutant alleles, thus losing much context detail (e.g. OMIM, [2]). mutation screening derived from ongoing CGP studies. Large datasets are therefore made available for deep datamining, whilst maintaining sample sizes which can still achieve good statistical significance.
The CGP maintains a listing of all genes which are
proved to have an involvement in causing cancer when
mutated, called the Cancer Gene Census (CGC,
http://www.sanger.ac.uk/genetics/CGP/Census/ [
Curation of the literature for 62 genes with oncogenic point mutations is now nearing completion (with the notable exception of TP53 which is independently curated; IARC TP53 database, [1]). Additionally, 15 genes have been assessed for their involvement in fusion events. Whilst the literature curation domain in COSMIC aims to completely capture the mutation data on a small set of known cancer genes resulting in large sample counts, the CGP domain aims to examine a small number of tumours through a large set of almost 4,800 candidate genes searching for novel oncomutations. COSMIC has now captured results on a total of 268,938 tumours which have been examined through 4,773 genes in various combinations representing 1,019,304 individual experiments. 5,902 publications, together with unpublished CGP contributions have described 59,187 small intragenic mutations and 2,266 instances of novel fusion mutations.
COSMIC's website aims to make searching, navigation
and interpretation as easy as possible by providing
much graphical summary with the contents of each
image reflected in a table with links to further details
(figure 1). Detailed descriptions of the system's
concepts, contents and usage have recently been
published [
The extraction of complex mutation annotation together
with clinical and phenotypic details about tumour
samples has proved beyond any retrieval method other
than manual. Building the COSMIC dataset has taken a
team of manual curators eight years and is still ongoing.
Attempts to automatically text-mine scientific
publications could retrieve simple data components, but
rarely captured the rich contextual information that
makes COSMIC so successful in cancer genetics.
With the advent of systematic candidate gene screening
and next generation sequencing technologies, the size
and scope of screens is increasing. In one of the largest
analysis to date, 22 tumours were examined through
over 18,000 genes in an attempt to find novel oncogenic
variants [
COSMIC's presentation of complex mutation data in a phenotypic context has proved very successful, with the website consistently registering around 400,000 page impressions per week in 2008. However, imparting meaning to each variant has become an immensely complex proposition, especially with the larger systematic screens examining anonymous genes of unknown function. Most of COSMIC makes very little distinction between mutations known to cause cancer and passenger mutations, unless this is discussed in the publication being curated, which is rare. Whilst it is easy to define, for instance, a frameshift mutation in a tumour suppressor gene as an oncogenic variant, it can be very difficult to determine the oncogenic consequence of a novel missense mutation. CGP's "Cancer Cell Line Project", displayed on COSMIC's green pages is the only part of the system to make this causal distinction. With highly unstable genomes, cell lines provide mutation hunters with many variants, many of which will have no significant oncogenic consequence. COSMIC's green pages therefore, only record and display those with obvious or previously characterised oncogenic characteristics; any variants not obviously oncogenic are not shown. How this consequence is measured, is again a manual process largely based upon previous published evidence that the variant has been observed somatic before or that it has clear functional consequences [7].
Many of the well known mutations in cancer have been
fully characterised; for instance, mutations at codons
12,13 and 61 in KRAS are known to interfere with the
binding of RAS and its inactivator, thus extending the
molecule's signalling, upregulating cellular growth via
the MAPK/ERK signalling cascade [8]. However,
missense mutations are recorded at over 20 other sites
in this gene and the meaning of these is much less clear,
even when the positions of these residues in the 3D
structure of the molecule is easily identifiable. Again,
software is available (e.g. SIFT [
1. Petitjean, A., Mathe, E., Kato, S., Ishioka, C., Tavtigian,S.V., Hainaut, P., Olivier, M. 2007. Impact of mutant p53 functional properties on TP53 mutation patterns and tumor phenotype: lessons from recent developments in the IARC TP53 database. Hum Mutat. 28:622-29. 2. Hamosh, A., Scott, A.F., Amberger, J., Bocchini, C., Valle, D., McKusik, V.A. 2002. Online Mendelian Inheritance in man (OMIM), a knowledgebase of human genes and genetic disorders. Nucl. Acids Res. 30:52-55. 6. Campbell P.J., Stephens P.J., Pleasance E.D., O'Meara S., Li H., Santarius T, Stebbings LA, Leroy C, Edkins S, Hardy C, Teague JW, Menzies A, Goodhead I, Turner DJ, Clee CM, Quail MA, Cox A, Brown C, Durbin R, Hurles ME, Edwards PA, Bignell GR, Stratton MR, Futreal PA. 2008. Nat Genet 40:685-686. 7. Ikediobi ON, Davies H, Bignell G, Edkins S, Stevens C, O'Meara S, Santarius T, Avis T, Barthorpe S, Brackenbury L, Buck G, Butler A, Clements J, Cole J, Dicks E, Forbes S, Gray K, Halliday K, Harrison R, Hills K, Hinton J, Hunter C, Jenkinson A, Jones D, Kosmidou V, Lugg R, Menzies A, Mironenko T, Parker A, Perry J, Raine K, Richardson D, Shepherd R, Small A, Smith R, Solomon H, Stephens P, Teague J, Tofts C, Varian J, Webb T, West S, Widaa S, Yates A, Reinhold W, Weinstein JN, Stratton MR, Futreal PA, Wooster R. Mutation analysis of 24 known cancer genes in the NCI-60 cell line set. Mol Cancer Ther. 2006, 5:260612. 8. Scheffzek K, Ahmadian MR, Kabsch W, Wiesmuller L, Lautwein A, Schmitz F, Wittinghofer A. The RasRasGAP complex:structural basis for GTPase activation and its loss in oncogenic RAS mutants. 1997. Science 277: 333-338.