Markus Ball

CG>TG mutation frequency as negative predictor of homologous recombination deficiency in ovarian and breast cancer

Abstract Homologous recombination deficiency (HRD) is a predictive biomarker for PARP inhibition and platinum-based chemotherapy. While copy number alteration-based scores such as HRDsum = LST + TAI + LOH are included in therapy approvals, single base substitutions (SBS) are underinvestigated as predictors of HRD. WES data of the TCGA pan-cancer cohort and an in-house ovarian cancer cohort were annotated by alterations in BRCA1/2 and additional genes causative of HRD. Using this reference, the new biomarker fdeam defined as frequency of C > T transitions at CpG sites in relation to all SBS and HRDsum were compared for the detection of HRD. In the TCGA ovarian cancer, the in-house, and the TCGA breast cancer cohorts, fdeam performed non-inferior to HRDsum (AUC = 0.84, AUC = 0.85, and AUC = 0.88). The cutpoint fdeam = 13.1% maximized the balanced accuracy in the TCGA ovarian cancer cohort and resulted in sensitivity = 89% and specificity = 77% in the in-house cohort. In a simulation study, fdeam retained high sensitivity for HRD detection and outperformed HRDsum in tumors of purity 40%, 20%, and 10%. Overcoming the limited robustness against low tumor purity, the new biomarker can contribute to a more sensitive detection of HRD in clinical samples. Further studies are warranted to confirm its clinical validity and utility and explore its potential for liquid biopsies.