Computational tools to detect signatures of mutational processes in DNA from tumours: a review and empirical comparison of performance
AbstractMutational signatures refer to patterns in the occurrence of somatic mutations that reflect underlying mutational processes. To date, after the analysis of tens of thousands of genomes and exomes from about 40 different cancers types, 30 mutational signatures characterized by a unique probability profile across the 96 mutation types have been identified, validated and listed on the COSMIC (Catalogue of Somatic Mutations in Cancer) website. Simultaneously with this development, the last few years saw the publication of several concurrent methods (mathematical algorithms implemented in publicly available software packages) for either the quantification of the contribution of prespecified signatures (e.g. COSMIC signatures) in a given cancer genome or the identification of new signatures from a sample of cancer genomes. A review about existing computational tools has been recently published to guide researchers and practitioners in conducting their mutational signatures analysis, however, other tools have been introduced since its publication and, to date, there has not been a systematic evaluation and comparison of the performance of such tools. In order to fill this gap, we carry on an empirical evaluation study of all available packages to date, using both real and simulated data.