Mutation bias in driver genes reveals the distribution of effects of oncogenic mutations

Abstract

Although many cancer driver genes have been identified, the full set of mutations within these genes that can promote tumorigenesis remains unknown. The contribution of a mutation depends on both its probability of occurrence and the selective advantage it confers. Here, we introduce a metric that quantifies how selection biases the spectrum of observed mutations in driver genes, and use it to develop a method estimating the number of driver mutations and their effect distribution at individual genes. Applying this framework to large cancer cohorts, we find that in most oncogenes, nearly all driver mutations have already been observed, whereas in most tumor-suppressor genes, the majority remain undiscovered. These results reveal fundamental differences in how mutation and selection shape mutational patterns across gene classes and provide a framework for interpreting newly detected variants.