An optimized homologous recombination deficiency (HRD) algorithm ASGAD for predicting PARP inhibitor response in ovarian cancer patient.

e17077 Background: Homologous recombination deficiency (HRD) results less efficient and error-prone DNA double strand break (DSB) repair, thus causes genomic in stability and impacts on cancer susceptibility to Poly‐(ADP‐Ribose)‐Polymerase (PARP) inhibitors. Evaluating HRD level in gynecologic cancer patients is becoming far more important and influential, so far, there is no standard method to be used in clinical. Methods: Here, we optimized an HRD score algorithm, termed as ASGAD, which combines three classic factors, including loss of heterozygosity score (LOH), telomeric allelic imbalance score (TAI), large-scale state transition score (LST), along with tumor ploidy to predict PARP inhibitor response in ovarian cancer patient. Results: Traditional global optimization strategy for purity and ploidy calculation is usually sensitive to initial values and may lead to incorrect convergences. Here we designed a two-step optimization manner to avoid this problem. Firstly, a density-based clustering algorithm was applied on BAF and CN, and the genotype was assigned to the most legible cluster. Then the rectified BAF and CN was calculated and used to find the maximum likelihood genotype of each segment. The segmentation processed was also improved by applying a series of statistical test to merge similar adjacent segments. The accuracy of allele-specific copy-number detection is significantly improved vie this algorithm, deriving stable and reliable HRD scores especially on aneuploid and hyperploid tumor cases. In this study, we assessed ASGAD algorithm in almost 150 ovarian cancer patient samples, who had treated with platinum effectively. BRCA1/2 deficient was defined as either one deleterious mutation in BRCA1/2, with LOH in the wild type copy or two deleterious mutations in the same gene. The results showed that the 19/23 BRCA1/2 deficient samples are also HRD-high, giving the sensitivity of 82.61%. Besides, we identify 58 HRD-high samples with intact BRCA1/2, who might benefit from PARP inhibitors, by the ASGAD algorithm. Interestingly, we analyzed 4 HRD-high samples by using whole-exome sequencing (WES), and found other HR genes mutations in these samples, including PARP4, FANCM, MSH2, MSH6, ATR, POLD1. Conclusions: The expanded use of PARP inhibitors in HRD tumors using the ASGAD algorithm requires more validation.