Direct Pan-Cancer Multi-Omic Analysis of Radiotherapy Response Reveals Coding and Regulatory Genomic Variation, and Extracellular Adaptive Programs Underlying Radioresistance

Radioresistance remains a major barrier to effective radiotherapy, yet its molecular determinants remain incompletely understood. Most radiogenomic studies infer radioresistance using surrogate endpoints rather than from clinically annotated treatment response data. Here, we analysed genomic, transcriptomic, immune and clinical features of 1,445 radiotherapy-treated patients across 15 cancer types represented in The Cancer Genome Atlas with documented radiotherapy outcomes. Cancer-type-specific somatic association analyses identified 132 significant candidate variant associations across six cancer types, including three previously unannotated coding variants in TCF20, PTPN14, and CADM1. Functional annotation showed that radioresistant tumours were enriched for cytoplasmic, protein-associated, and metabolic functions, whereas radiosensitive tumours were enriched for nuclear regulatory programs. Through transcriptomic analysis, we identified 107 differentially expressed genes across 8 cancer types, enriched for extracellular signalling, secretory programs, and tumour microenvironment interactions. The genomic and transcriptomic analyses revealed a limited gene overlap but converged on cellular adaptation and survival. Together, our findings suggest that radioresistance is characterised by complementary adaptive processes, whereby intracellular protein-regulatory and metabolic processes are coupled with extracellular signalling and tumour microenvironment interactions. Clinically, radioresistance was associated with significantly poorer overall survival, whereas the radiosensitivity index and inferred tumour immune composition showed limited concordance with observed treatment response. To our knowledge, this study provides the first pan-cancer characterisation of clinically defined radiotherapy response and identifies potential candidate biomarkers for functional validation.