Abstract LB-401: Functional impact bias reveals cancer drivers

Abstract

Abstract Currently, one of the main challenges in cancer genomics is the identification of driver genes and pathways among all the altered genes in a cohort of tumors. Typically, tests of recurrence are applied to identify significantly mutated genes, which are candidates to cancer drivers. These methods assess the probability to observe by chance the number of somatic single nucleotide variants (SNVs) found across a number of tumor samples. Some known limitations of these methods include the difficulty in correctly assessing background mutation rate, as all parameters that affect it are not well-understood, and the fact that they usually fail to identify lowly recurrently mutated driver genes. In addition, methods that compute recurrence treat all observed somatic SNVs equally when clearly their impact on protein function varies widely. We have developed a method which can complement the analysis of recurrence of SNVs across several tumor samples to detect candidate driver genes by taking advantage of the information provided by the analysis of functional impact of SNVs and short frameshift indels. The method is based on the assumption that any bias towards the accumulation of variants with high functional impact (FM bias) is an indication of positive selection and can thus be used to detect candidate driver genes or gene modules. We have called it Oncodrive-fm. We have applied the Oncodrive-fm approach to two datasets of genes with variants in samples of different tumor types: Chronic Lymphocytic Leukemia (cll: from Spanish ICGC Group) and Glioblastoma datasets (gbm: from TCGA). Most of the genes detected by recurrence analysis (i.e MutSig) also show significant FM bias (eg. TP53, EGFR, PTEN in gbm and SF3B1 in cll), we call these clear drivers. However, Oncodrive-fm can also detect lowly recurrent likely driver genes that are overlooked by recurrence analysis (eg. IDH1, FGFR1 and FGFR3 in gbm and CDH9 and CHD2 in cll), and identify genes with high number of mutations that do not show FM bias, which we term as candidate false positives. The application of Oncodrive-fm to the analysis of pathways reveals several pathways with FM bias, such as MAPK and mTOR pathways in gbm and mRNA splicing pathway in cll. In summary we show that FM bias helps to uncover drivers and recommend its computation in conjunction with the analysis of recurrence in a cohort of tumor samples to identify candidate driver genes and gene modules. Oncodrive-fm method is available at http://bg.upf.edu/oncodrive. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr LB-401. doi:1538-7445.AM2012-LB-401