Abstract 1510: Subtype-specific cancer driver gene detection improves sensitivity to detect drivers

Abstract Identifying cancer driver mutations is a crucial step toward understanding the underlying mechanisms of oncogenesis. However, driver gene detection is complicated, considering the inherent complexity and heterogeneity of cancer.The past decade has seen the large-scale application of next-generation sequencing technologies (NGS) in cancer genomics, and many methods have been published utilizing NGS to detect novel oncogenic drivers. However, the low concordance across methods raises the concern about false positives and negatives in those findings. In particular, proper modeling of the background mutation profile (known to be affected by multiple factors) is critical to improving precision and recall in driver detection. Though intra-sample heterogeneity in the background mutation profile is considered to some extent in multiple methods, most computational driver detection methods assume a homogeneous mutational landscape across cancer samples. With more and more subtypes being discovered in various cancers, the assumption is largely untenable. In this study, we present a driver gene detection framework taking into account the heterogeneous mutational context in a cancer cohort. Our approach improves sensitivity to detect drivers by first pre-selecting a more uniform sample subset to apply driver detection algorithms. We combine this method with enhancements to an existing ensemble approach that combines methods with different assumptions about the characteristics of driver mutations (recurrence across samples, functional impact bias and positional clustering). We apply this combined approach to ∼750 breast invasive carcinoma samples from The Cancer Genome Atlas. In this dataset, genes identified by our approach show higher enrichment in known cancer driver genes compared to other methods, and the subtype-driven method identifies additional potential drivers that are missing in overall analysis. Citation Format: Yao Fu, Aparna Chhibber, Narges Bani-Asadi, Hugo YK Lam. Subtype-specific cancer driver gene detection improves sensitivity to detect drivers. [abstract]. In: Proceedings of the 107th Annual Meeting of the American Association for Cancer Research; 2016 Apr 16-20; New Orleans, LA. Philadelphia (PA): AACR; Cancer Res 2016;76(14 Suppl):Abstract nr 1510.