Learning mutational signatures and their multidimensional genomic properties with TensorSignatures

Abstract We present TensorSignatures, an algorithm to learn mutational signatures jointly across different variant categories and their genomic localisation properties. The analysis of 2778 primary 3824 metastatic cancer genomes the PCAWG consortium HMF cohort shows that all operate dynamically in response states. pins differential spectra UV mutagenesis found active inactive chromatin global genome nucleotide excision repair. TensorSignatures accurately characterises transcription-associated 7 types. also extracts distinct replication- double strand break repair-driven by APOBEC3A 3B with numbers length mutation clusters. Finally, reproduces a signature somatic hypermutation generating highly clustered variants at transcription start sites genes lymphoid leukaemia, from general less Pol?-driven translesion synthesis broad range In summary, elucidates complex footprints characterising underlying processes respect multitude variables.