This session will highlight biological insights and potential implications revealed by HRDsig across tumor types, including ovary, breast, pancreas, prostate, lung, and tumor-agnostic settings. HRDsig is a machine learning–based genomic scar biomarker that quantifies genome-wide copy number features to detect homologous recombination deficiency (HRD) across tumor types. By capturing the downstream genomic consequences of somatic or germline homologous recombination repair (HRR) gene alterations and epigenetic silencing events, HRDsig identifies HRD in both HRR-mutated and wild-type tumors. Retrospective pan-cancer analytical validation analyses demonstrated high sensitivity, specificity, and reproducibility across diverse samples, including those of borderline quality. HRDsig score has been evaluated in more than 500,000 samples and has been shown to increase both sensitivity and specificity for the detection of HRD.
