SSA.ME Detection of cancer mutual exclusivity patterns by small subnetwork analysis

Abstract

Because of its clonal evolution a tumor rarely contains multiple genomic alterations in the same pathway, as disrupting the pathway by one gene often is sufficient to confer the complete fitness advantage. As a result mutated genes display patterns of mutual exclusivity across tumors. The identification of such patterns have been exploited to detect cancer drivers. The complex problem of searching for mutual exclusivity across individuals has previously been solved by filtering the input data upfront, analyzing only genes mutated in numerous samples. These stringent filtering criteria come at the expense of missing rarely mutated driver genes. To overcome this problem, we present SSA.ME, a network-based method to detect mutually exclusive genes across tumors that does not depend on stringent filtering. Analyzing the TCGA breast cancer dataset illustrates the added value of SSA.ME: despite not using mutational frequency based-prefiltering, well-known recurrently mutated drivers could still be highly prioritized. In addition, we prioritized several genes that displayed mutual exclusivity and pathway connectivity with well-known drivers, but that were rarely mutated. We expect the proposed framework to be applicable to other complex biological problems because of its capability to process large datasets in polynomial time and its intuitive implementation.