A real-world pan-cancer catalogue of mutational signatures from 111,711 tumors

Tumor mutational signatures record the genomic footprints of endogenous and exogenous mutagenic processes and can refine cancer diagnosis and treatment. However, their clinical translation has been limited because most reference catalogues and fitting algorithms are calibrated for whole-exome or whole-genome sequencing, whereas routine diagnostics often use targeted sequencing panels that sample uneven genomic regions. Here we introduce SATS, a computational framework designed for mutational signature analysis in targeted sequencing data. Applying SATS to 111,711 tumors from the AACR Project GENIE, we generated a real-world, panel-calibrated pan-cancer catalogue of mutational signatures for targeted sequencing and evaluated the reproducibility of lung, breast and colorectal signatures in 22,388 additional tumors from a later GENIE release and three external cohorts. Because these tumors were profiled in routine clinical care and linked to clinical records, the resulting catalogue is anchored in a real-world oncology setting rather than in a research-sequencing cohort. The catalogue spans a broad spectrum of cancer types and subtypes, revealing pervasive heterogeneity in signature prevalence. Integrating this catalogue with matched clinical records, we identified signatures enriched in early-onset hypermutated colorectal cancer and signatures associated with cancer prognosis and response to immunotherapy. Together, SATS and this catalogue provide a mutational signature resource for studying cancer etiology, prognosis and treatment response from clinical targeted sequencing.