PROJECT SUMMARY/ABSTRACT All genetic variation—including that underlying heritable disease, cancer, and human evolution—originate from mutation. Recent large-scale genome sequencing efforts and innovative statistical analysis have uncovered substantial variation in mutation rate along the human genome, revealing strong impacts of the mutation type and flanking sequence. However, the molecular mechanisms of this context-dependency of mutation rate are poorly understood. In addition, mutation rate variation across genomic sites may bias inferences of selection signals from genomic data, which in turn hinders the identification and functional study of genes that drive disease. The goal of our research program is to develop computational methods to draw insights into the molecular mechanisms as well as functional and evolutionary consequences of context-dependent mutation rate variation. We will first focus on the hypermutability of CpG sites and take a multifaceted approach to investigate the lesion formation and repair at methylated cytosines in different regions of the genome, by utilizing existing genomic and epigenomic data from human populations and other species. Successful completion of this research will contribute to a mechanistic and quantitative understanding of the mutational processes at methylated cytosines. Next, we will improve computational methods for inferring selection on human genes by leveraging the inherent mutation rate variation across genomic sites. We propose to combine population genetics models and machine learning techniques to integrate the allele frequency, site-specific mutation rate, and functional information of variants. The application of these newly developed methods to the ever-growing genomic data will identify genes crucial to human health and reproduction, and improve estimates of the burden of deleterious variants introduced by new mutations in each generation. Finally, we will expand our research scope to somatic mutations and leverage the context-dependency of mutation rates to better understand cancer driver genes. We will evaluate the relative mutability of cancer driver genes under different mutational processes, and investigate how tissue-specific relative mutability and selective effect interact during somatic evolution of tumor. By taking an evolutionary perspective of tumorigenesis, this research promises to shed new light on the tissue-specificity of cancer driver genes. Together, the proposed research will develop novel computational approaches that translate the rich genomic and epigenomic data available into insights into mutational mechanisms, as well as selective forces acting on genes and genetic variants in human populations and somatic cells.