Clinical and genomic characterization of mutational signatures across human cancers

Mutational signatures, the generic patterns of mutations, are the footprints of both endogenous and exogenous factors that have influenced cancer development. To date, dozens of mutational signatures have been discerned through computational methods. However, the etiology, mutational properties, clonality, immunology and prognostic value of mutation signatures across cancer types are poorly understood. To address this, we extensively characterized mutational signatures across 8836 cancer samples spanning 42 cancer types. We confirmed and extended clinical and genomic features associated with mutation signatures. Mutation distribution analysis showed that most mutation processes were depleted in exons and APOBEC signatures (SBS2 and SBS13), the Pol-η related signature (SBS9) and SBS40 tended to contribute clustered mutations. We observed that age-related signatures (SBS1 and SBS5) and SBS40 tended to induce mutations affecting cancer genes and subclonal drivers posted by specific signatures (eg, mismatch repair deficiency-related signature SBS44) were unlikely subjected to positive selection. We also revealed early mutation signatures (eg, UV light exposure-related signature SBS7a) and signatures (eg, reactive oxygen species-related signature SBS18) predominated in the late stage of tumorigenesis. Comprehensive association analysis of mutation processes with microenvironment revealed that APOBEC- and mismatch repair deficiency-related signatures were positively associated with immune parameters, while age-related signatures showed negative correlations. In addition, prognostic association analysis showed that many signatures were favorable (eg, SBS9) or adverse factors (eg, SBS18) of patient survival. Our findings enhance appreciation of the role of mutational signatures in tumor evolution and underline their potential in immunotherapy guidance and prognostic prediction.