Preprint on mutational processes in cancer genomes

A new preprint on regional mutational processes in cancer genomes is now available in bioRxiv. The study investigated hundreds of epigenetic profiles of human cancers as predictors of megabase-scale mutation frequencies in 2500 whole cancer genomes using machine learning with random forests. We found that epigenetic profiles of cancer samples, rather than those normal samples, are the strongest predictors of mutagenesis and mutational signatures in most cancer types, while excess mutations unexplained by these epigenetic profiles are enriched in cancer driver genes and various molecular processes. The study builds on the Pan-cancer Analysis of Whole Genomes (PCAWG) project and was led by Oliver Ocsenas, a recent MSc graduate of the Medical Biophysics Department of University of Toronto.