Abstract 2483: DirectHRD enables sensitive scar-based classification of homologous recombination deficiency (HRD)

Introduction: Homologous recombination deficiency (HRD) is a predictive biomarker for response to PARP inhibition and platinum chemotherapy but remains challenging to detect. The discovery of patients without pathogenic mutations in known HR genes but exhibiting genomic scars indicative of HRD led to the FDA approval of the first scar-based HRD test. Despite advancements in whole genome sequencing (WGS) combined with machine learning models, current methods lack the sensitivity required for detecting HRD scars in low tumor purity samples such as liquid biopsies. Method: We developed DirectHRD, a WGS-based HRD scar classifier. Compared to other HRD scar classifiers, DirectHRD exclusively utilizes a highly specific HRD scar—deletion with microhomology—and its associated signatures (COSMIC ID6 signature) in a probabilistic framework to detect HRD scars with higher sensitivity. Results: We first applied DirectHRD to 474 tumor biopsy samples from the PCAWG (Pan-Cancer Analysis of Whole Genomes) project, focusing on the four cancer types with the highest HRD prevalences (breast, ovary, pancreas, and prostate). We achieved 100% detection of patients with bi-allelic BRCA1/2 losses with high specificity (>90%) across all four cancer types. Next, we applied DirectHRD to WGS of tumor biopsies and in-silico dilutions from 27 patients with stage II-III triple-negative breast cancer from the TBCRC030 trial. DirectHRD achieved an AUC of 0.99 at 5% tumor fraction, compared to an AUC of 0.62 and 0.48 for CHORD and HRDetect, respectively. Finally, we applied DirectHRD to 90 ctDNA (circulating tumor DNA) samples across three cancer types (breast, ovarian, and prostate). The ground truth HRD statuses were determined from tumor biopsies using scar-based methods for breast (Myriad MyChoice, scarHRD and SigMA) and ovarian cancers (scarHRD and SigMA) and by identifying germline and somatic pathogenic mutations in HR genes for prostate cancers. DirectHRD achieved an AUC of 0.86 in stage II-III breast cancers and 0.9 in metastatic breast cancers. For ovarian cancers and metastatic prostate cancers, DirectHRD achieved AUCs of 0.9 and 0.87, respectively. Conclusion: Our findings suggest that DirectHRD can reliably detect HRD scars in low tumor purity samples, such as liquid biopsies, using moderate depths (e.g., 50x) of WGS. This could broaden access to HRD testing including when tissue biopsies are insufficient or infeasible. Citation Format: Ruolin Liu, Eugenia Roberts, Heather Parsons, Elizabeth Stover, Atish Choudhury, Justin Rhoades, Timothy Blewett, David Yang, Joyce Liu, Erica Mayer, Viktor Adalsteinsson. DirectHRD enables sensitive scar-based classification of homologous recombination deficiency (HRD) [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 2483.