Characterizing selection signatures in coding and noncoding regions of 14,886 cancer genomes

Clonal selection drives cancer development, but quantifying selection on noncoding somatic mutations remains largely unexplored. Here, we introduce dNdS-Fun, an extension of the dN/dS framework to quantify selection of both coding and noncoding somatic mutations, thereby enhancing the discovery of driver genes. Applying dNdS-Fun to whole-genome sequencing data from 14,886 cancer patients across 31 cancer types, we identified 196 genes under positive selection across multiple cancer types or datasets. Of these, 83 were previously unrecognized as drivers, and 56 were identified solely through noncoding mutations. Additionally, we observed widespread negative selection throughout the genome, particularly enriched in essential or cancer-dependent genes. Twenty genes exhibited an overall signature of negative selection but showed positive selection in noncoding elements, indicating both their conserved functions and adaptive regulatory roles in tumorigenesis. Our study reveals pervasive selection signatures of non-coding mutations, providing important insights for future research on their roles in cancer progression.