BackgroundHomologous recombination deficiency (HRD) score serves as a promising biomarker to identify patients who are eligible for treatment with PARP inhibitor (PARPi). Previous studies have suggested a 3-biomarker Genomic Instability Score (GIS) threshold of ≥ 42 is a valid biomarker to predict response to PARPi in patients with ovarian cancer and breast cancer. However, the GIS threshold for prostate cancer (PCa) is still lacking. Here, we conducted an exploratory analysis to investigate an appropriate HRD score threshold and to evaluate its ability to predict response to PARPi in prostate cancer patients. MethodsA total of 181 patients with metastatic castration-resistant PCa were included in this study. Tumor tissue specimens were collected for targeted next-generation sequencing for homologous recombination repair (HRR) genes and copy number variation (CNV) analysis. HRD score was calculated based on over 50,000 single-nucleotide polymorphisms (SNP) distributed across the human genome, incorporating three SNP-based assays: loss of heterozygosity, telomeric allelic imbalance, and large-scale state transition. The HRD score threshold was set at the last 5th percentile of the HRD scores in our cohort of known HRR-deficient tumors. The relationship between HRD score and efficacy in 16 patients of our cohort who received PARP inhibitors were retrospectively analyzed. ResultsGenomic testing was succeeded in 162 patients. In our cohort, 61 patients (37.7%) had HRR mutations (HRRm). BRCA mutation occurred in 15 patients (9.3%). The median HRD score was 4 (ranged from 0 to 57) in total cohort which is much lower than that of in breast and ovarian cancer. Patients harbored HRRm, BRCA or TP53 mutation have a higher HRD score. CNV occurs more frequently in patients with HRRm. The last 5th percentile of HRD scores was 43 in the HRR-mutant cohort and consequently HRD high was defined as HRD scores ≥43. In the 16 patients who received PARPi in our cohort, 4 patients with a high HRD score achieved an objective response rate (ORR) of 100% while 12 patients with a low HRD score achieved an ORR of 8.3%. Progression-free survival (PFS) in HRD high patients is longer compared to HRD low patients, regardless of HRRm. ConclusionsA HRD score threshold of 43 is established and preliminarily validated to predict the efficacy of PARPi in this study. Future studies are needed to further verify this threshold.