Genomic foundation model-derived disruption profiling links somatic mutations to cancer biology and clinical outcomes

The research examined how somatic mutations in DNA can cause partial gene disruption with biological and clinical consequences. Scientists used AlphaGenome and AlphaMissense models to analyze mutations in 8,800 patients with 33 cancer types from The Cancer Genome Atlas database. They found that recurrent hotspot mutations had substantially larger predicted effects at the protein level, while other mutations showed greater regulatory effects. They created patient-level gene disruption profiles reflecting tissue type, cancer type, and microsatellite instability status. Among patients lacking recurrent hotspot mutations in a specific cancer gene, higher predicted disruption was associated with overall survival, with the strongest signal observed in chromatin accessibility. The results indicate that gene disruption has a continuous, multidimensional character extending beyond discrete cancer drivers.