Abstract 1682: Uncovering signals of somatic selection through whole exome and whole genome sequencing of lung adenocarcinoma

Abstract

Abstract We have sequenced the exomes of over 100 and the genomes of over 20 lung adenocarcinoma tumor-normal specimen pairs. We performed hybrid capture exome sequencing of nearly 18,000 genes to >100X median per-sample coverage with 76bp paired-end reads. We performed whole genome sequencing achieving 60X median tumor and 20X median normal coverage with 350bp median insert size and 101bp paired-end reads. Our exome analysis yielded over 50,000 substitution and small indel coding events, with a mean somatic mutation rate of 10-11 events / MB. This resulted in over 300 non-synonymous coding events per patient, most of which were presumed to be passenger mutations unrelated to tumorigenesis. This high mutational load required us to develop novel statistical approaches (MutSig, Lawrence et al, in preparation) to identify putative lung adenocarcinoma driver genes under positive somatic selection. We constructed a complex neutral mutation model that considered sequence context and several additional genomic covariates shown to mediate gene to gene inhomogeneities of passenger mutation rates. Identification of significant deviations from this background model allowed us to recover almost all known frequently mutated lung adenocarcinoma genes, including TP53, KRAS, STK11, PIK3CA, EGFR, ERBB2, RB1, SMARCA4, and KEAP1, as well as a host of novel putative driver genes. We applied similar principles to identify pathways and sub-networks of genes undergoing apparent positive selection in lung adenocarcinoma. Whole genome analysis yielded several high-confidence in-frame protein fusion and promoter-gene fusion events enriched in tumor vs normal specimens. We also found large numbers of somatic substitution and indel events in promoters, enhancers, and non-coding DNA elements and identified putative sites of somatic retrotransposition in our whole genome data. Overall, our study eclipses previous large-scale characterization of somatic sequence variation (Refs. 1-3) in lung adenocarcinoma by at least an order of magnitude. Using novel methods adapted to the analysis of high-mutation rate tumor types (lung squamous cell carcinoma, melanoma, colorectal cancer), we are able to recover signals of selection in both known and novel genes and pathways. Our results illuminate novel lung adenocarcinoma tumor biology and provide targets for therapeutic and diagnostic investigation.