DoCM: a database of curated mutations in cancer

Large-scale cancer genomics discovery projects, such as The Cancer Genome Atlas (TCGA), and the International Cancer Genome Consortium (ICGC) among others, have systematically characterized the molecular lesions in human cancer genomes, thereby laying the foundation for precision cancer medicine. However, a curated set of somatic variants with established relevance to cancer biology is essential for clinical annotation and for use in computational data analysis. We have created DoCM, a Database of Curated Mutations in cancer (http://docm.info), as an open-source, openly-licensed resource to enable the cancer research community to aggregate, store, and track biologically important cancer variants.

A variety of somatic cancer variant databases exist that help identify important variants, including gene-level ¹, variant-level ^2,3, and clinically-focused variant interpretation databases ^4-6. These resources have greatly increased our understanding of the landscape of clinically and biologically relevant cancer variants, and when used in aggregate provide an understanding of the relevance of specific variants. DoCM is a curated repository that facilitates the aggregation of gene/variant information for variants with prognostic, diagnostic, predictive or functional roles from these resources as well as individually curated publications (Supplementary Figure 1, Supplementary Table 1). The data model and batch submission process (Supplementary Results, Supplementary Figures 2-4) used by DoCM places it at a critical intersection between the two major tradeoffs of curated resources: comprehensiveness of variants and curation burden (Figure 1). In a rapidly changing landscape of genes and variants for which new information is steadily accumulating, an automated batch submission and review system allows DoCM curations to scale easily.

Figure 1. DoCM supports existing curation initiatives while occupying a critical niche that balances comprehensiveness and curation burden.

DoCM accepts variant batch submission of arbitrary size and varying complexity, allowing the resource to be agile and comprehensive. The DoCM data model limits curation burden, while permitting the entry of genes and variants with high quality functional data. DoCM also aggregates functionally important variants from many other quality resources. CIViC, a knowledgebase of clinical interpretations of variants in cancer (http://civicdb.org), is focused on summarizing and aggregating evidence of clinically actionable variants into clinical interpretations. ClinVar aggregates structured variant records and clinical assertions, but has largely been focused on germline variants. Variant observation databases, like ICGC, COSMIC, and TCGA, attempt to report the totality of somatic variants observed in patients to-date. All of these databases are complimentary and inform each other.

Curation of the literature to produce a high quality set of pathogenic somatic variants is not trivial and it is unrealistic that one group could independently keep pace with the ever-expanding cancer genomics literature (Supplementary Figure 5). Hence, we have designed DoCM as an open resource that can coordinate contributions from research and clinical practitioners in cancer genomics. Once important variants are identified, they require significant curation efforts to format and standardize the variants in a structured way for storage and retrieval in a relational database (Supplementary Methods, Supplementary Figure 6). A set of such curated variants can be contributed to DoCM by batch submission at http://docm.genome.wustl.edu/variant_submission, whereupon they are reviewed and evaluated by DoCM editors for possible inclusion. DoCM is licensed under the creative commons attribution license (CC BY 4.0), allowing academic and industry researchers to freely access the content.

DoCM provides easy access to a current and accurate list of functionally important cancer variants with clear provenance, based on peer-reviewed journal citations. The content of DoCM may be accessed via a web interface or a documented application programming interface (API). To illustrate the utility of DoCM, we performed a focused knowledge-based variant discovery study to identify pathogenic variants missed in 1,833 cases across four TCGA projects (Supplementary Methods, Supplementary Figure 7). Validation sequencing data from 93 of these cases showed that at least one functionally important variant in DoCM was recovered in 41% of cases (Supplementary Results, Supplementary Data 1-2, Supplementary Figure 7-9, Supplementary Table 2-4). As genomics evolves into the era of precision medicine and our understanding of the etiology of molecular lesions grows, community curation along with our ongoing efforts will allow DoCM to adapt, refine, and expand with the field.