Detecting HRD in whole-genome and whole-exome sequenced breast and ovarian cancers

• Published in: medRxiv : the preprint server for health sciences
• Main Authors: Abbasi, Ammal, Steele, Christopher D, Bergstrom, Erik N, Khandekar, Azhar, Farswan, Akanksha, Mckay, Rana R, Pillay, Nischalan, Alexandrov, Ludmil B
• Format: Journal Article
• Language: English
• Published: United States 14.07.2024
• DOI: 10.1101/2024.07.14.24310383

Breast and ovarian cancers harboring homologous recombination deficiency (HRD) are sensitive to PARP inhibitors and platinum chemotherapy. Conventionally, detecting HRD involves screening for defects in , , and other relevant genes. Recent analyses have shown that HRD cancers exhibit characteristic mutational patterns due to the activities of HRD-associated mutational signatures. At least three machine learning tools exist for detecting HRD based on mutational patterns. Here, using sequencing data from 1,043 breast and 182 ovarian cancers, we trained Homologous Recombination Proficiency Profiler (HRProfiler), a machine learning method for detecting HRD using six mutational features. HRProfiler's performance is assessed against prior approaches using additional independent datasets of 417 breast and 115 ovarian cancers, including retrospective data from a clinical trial involving patients treated with PARP inhibitors. Our results demonstrate that HRProfiler is the only tool that robustly and consistently predicts clinical response from whole-exome sequenced breast and ovarian cancers. HRProfiler is a novel machine learning approach that harnesses only six mutational features to detect clinically useful HRD from both whole-genome and whole-exome sequenced breast and ovarian cancers. Our results provide a practical way for detecting HRD and caution against using individual HRD-associated mutational signatures as clinical biomarkers.