Characterizing Multimodal Mutational Processes in Cancer Using MOTET

Mutational signatures are characteristic patterns of somatic alterations in cancer genomes. A single aberrant mechanism can simultaneously generate coordinated patterns of alterations across multiple mutation modalities, including single-base substitutions (SBSs), doublet-base substitutions (DBSs), and small insertions and deletions (IDs). However, these signatures are typically inferred separately and linked post hoc, obscuring their shared underlying mechanisms. To resolve this, we formally define a Mutational Process (MP) as a unified latent activity that induces coupled mutational signatures across these modalities. We developed MOTET, a Bayesian framework that jointly infers multimodal MPs. Applied to 2,583 whole genomes from the PCAWG cohort, MOTET recovers known MPs more accurately than standard post hoc correlation methods. MOTET also identifies a process linking an SBS37-related signature with a TG>CA-dominated DBS component and resolves canonical SBS12-like activity into two separate MPs with distinct tissue distributions. These results establish a paradigm for formally defining catalogs of MPs and demonstrate that integrating across modalities enhances the ability to discover new mutational patterns.