Clinical validation of a comprehensive genomic profiling–based approach for homologous recombination deficiency assessment.

Abstract

e15087 Background: Homologous recombination deficiency (HRD) is a clinically actionable genomic biomarker in ovarian cancer and other solid tumors, guiding therapeutic selection and translational research. HRD testing is traditionally performed using dedicated assays; however, comprehensive genomic profiling (CGP) platforms offer the opportunity to integrate HRD assessment within routine molecular testing workflows. We evaluated the concordance and diagnostic performance of a CGP-based HRD prediction approach using clinically and genomically characterized tumor cohorts. Methods: A total of 122 ovarian cancer samples (111 tissue biopsy + 11 liquid biopsy samples) and 74 pan-cancer samples (10 Breast cancer, 57 Prostate cancer, and 7 pancreatic cancer) were retrospectively analyzed using OncoIndx comprehensive gene panel sequenced on NextSeq 550 Dx platform. HRD scores were determined from genome-wide patterns of loss of heterozygosity (LOH), large-scale transitions (LST), and telomeric allelic imbalance (TAI). In the external ovarian cancer cohort, HRD status was benchmarked against Myriad myChoice HRD scores, while BRCA1/2 stratification was used as a reference standard in the internal cohort. Diagnostic performance was assessed using sensitivity, specificity, predictive values, and overall accuracy. Concordance between tissue and liquid biopsy HRD status was evaluated in paired samples. Results: In the ovarian cancer cohort, CGP-based HRD predictions demonstrated strong concordance with myChoice classifications, achieving approximately 95% sensitivity and overall diagnostic accuracy exceeding 90%. In the BRCA-stratified cohort, HRD-high predictions showed high alignment with reference classifications, with sensitivity and specificity near 90% and diagnostic accuracy above 90%. Performance remained consistent across combined datasets, supporting stable and reproducible HRD prediction. In liquid biopsy ovarian cancer samples, HRD classification achieved 93% overall accuracy, with 88% specificity, 67% sensitivity. In paired pan-cancer tissue-liquid biopsy samples, HRD status demonstrated 73% concordance. Conclusions: This CGP-based HRD prediction approach enables simultaneous detection of actionable mutations and HRD status from a single assay and demonstrates high concordance with established HRD reference standards across ovarian and pan-cancer cohorts, supporting streamlined clinical testing and expanded access to HRD-informed therapeutic decision-making in both tissue and liquid biopsy settings.