Development and Validation of a Genome-Wide Association Study Based Polygenic Risk Score for Prostate Cancer in an Asian Population

Purpose: This study aimed to estimate genetic susceptibility to prostate cancer (PCa) by constructing a polygenic risk score (PRS) using single nucleotide polymorphisms (SNPs) identified from genome-wide association studies. Materials and Methods: The study included 1,015 PCa patients from our institutions and 1,015 age-matched controls from the Taiwan Biobank (TWB). An independent external validation cohort of 188 PCa patients and 188 TWB controls (excluding those from the primary cohort) was assembled. DNA was extracted from blood samples, with approximately 690,000 SNPs genotyped (minor allele frequency >= 0.05) and 15 million additional SNPs imputed using the 1000 Genomes Project. After quality control, 958 PCa patients and 999 controls were included in the analysis. The PRS was developed using PRSice2 by dividing samples into a base dataset and a model-testing set. Model performance was assessed using receiver operating characteristic analysis and cross-validation (CV). Results: Of the 87,092 SNPs initially considered, 24 were used to construct the PRS, located in intronic regions of genes such as KCNH7, HLA-DQA1, and PRNCR1. The PRS significantly improved PCa prediction, achieving an area under the curve (AUC) of 0.824 (p=1.23x10-50). Patients in the top 25th percentile of PRS had a 34-fold higher risk compared to those in the bottom 25th percentile (odds ratio=34.37, 95% confidence interval=22.93-51.68, p=1.96x10-52. The model showed stable performance with mean accuracies of 0.75 (3-fold CV) and 0.76 (10-fold CV) and achieved an AUC of 0.757 in the independent validation cohort. Conclusions: The developed PRS showed robust predictive ability for PCa in the Taiwanese population and may inform future risk stratification and personalized interventions.