Discovering and Categorizing Mutational Signatures From Cancer Whole Genomes

Mutational signatures are patterns of somatic mutations found in cancer genomes. They indicate past exposure to mutational processes, which can be either exogenous, such as DNA damage from UV light and chemicals in cigarettes, or endogenous, such as defects in DNA repair. We have analysed 18,640 cancer whole genomes across 21 tissue-types from Genomics England, ICGC and Hartwig Medical Foundation, uncovering 40 new single base substitution and 18 new double base substitution signatures. The analysis revealed that mutational signatures show tissue-specificity and can be either common or rare. Given a certain tissue-type, common signatures are present in most samples, with a median of five common signature for each sample, while rare signatures are present in only 5-15% of samples, with usually at most one rare signature per sample. Following these observations, we designed a signature fit algorithm, FitMS, which, given a new cancer sample and its tissue-type, can identify its common and rare signatures automatically.