Pervasive conditional selection of driver mutations and modular epistasis networks in cancer

Cancer driver mutations often display mutual exclusion or co-occurrence, underscoring the key role of epistasis in carcinogenesis. However, estimating the magnitude of epistatic interactions and their quantitative effect on tumor evolution remains a challenge. We developed a method to quantify COnditional SELection on the Excess of Nonsynonymous Substitutions (Coselens) in cancer genes. Coselens infers the number of drivers per gene in different partitions of a cancer genomics dataset using covariance-based mutation models and determines whether coding mutations in a gene affect selection for drivers in any other gene. Using Coselens, we identified 296 conditionally selected gene pairs across 16 cancer types in the TCGA dataset. Conditional selection accounts for 25-50% of driver substitutions in tumors with >2 drivers. Conditionally co-selected genes form modular networks, whose structures challenge the traditional interpretation of within-pathway mutual exclusivity and across-pathway synergy, suggesting a more complex scenario, where gene-specific across-pathway interactions shape differentiated cancer subtypes.

Pairwise tests for conditional selection use evolutionary logic to predict intrinsic drug resistance in ALK alterations

BackgroundGenomic data can facilitate personalized treatment decisions by enabling therapeutic hypotheses in individual patients. Conditional selection (encompassing both mutual exclusivity and co-occurrence) is commonly used to consider rare genomic variants relative to established cancer drivers. However, the direct comparison of p-values across multiple pairs of genes is confounded by the often-dramatic differences in sample size between established driver mutations and novel findings.MethodsWe develop a resampling based method for the direct pairwise comparisons of conditional selection between sets of gene pairs. This effectively creates quantitative positive control guideposts of conditional selection that normalize differences in population size. We applied this method to a transcript variant of ALK in melanoma, termed ALKATI, which has been the subject of a recent controversy in the literature. We reproduced some of the original cell transformation experiments, performed rescue experiments, and analyzed drug response data to revisit the original ALKATI findings.FindingWe found that we are powered to quantitatively compare the degree of relative mutual exclusivity (down to an abundance of 10 patients in a cohort of 500) between novel gene variants and positive controls. We also found that ALKATI is not as mutually exclusive as BRAF and NRAS are with each other. Our in vitro transformation assays and rescue assays suggested that alternative transcript initiation in ALK is not likely to be necessary or sufficient for cellular transformation or growth.InterpretationPairwise comparisons of conditional selection represent a sensitive method of utilizing existing genomic data to directly and quantitatively compare relative levels of conditional selection. The results of using our method in ALKATI and our experiments strongly disfavor the role of ALKATI as a targetable oncogenic driver. The progress of other experimental agents in late stage melanoma and our experimental and computational re-analysis led us to conclude that further single agent testing of ALK inhibitors in patients with ALKATI should be limited to cases where no other treatment hypotheses can be identified.