Feed

Early-stage diagnosis of paroxysmal atrial fibrillation (PAF) is challenging owing to its asymptomatic nature. However, the genetic factors underlying PAF and predictive utility of polygenic risk scores (PRSs) for PAF in Asian populations remain elusive. We aimed to explore the PAF-associated genetic variants in a Japanese cohort and evaluate the predictive performance of PAF-specific PRSs. This study included 2,604 participants. Following exclusion, quality control, and genotype imputation, a genome-wide association study (GWAS) was conducted. The predictive performance of 30 sets of PRS models constructed across various thresholds was evaluated using three machine learning methods. Model performance was assessed using area under the curve (AUC) and SHapley Additive exPlanations (SHAP). The GWAS using 1,038 PAF cases and 744 controls identified 82 genome-wide significant variants (P < 5 × 10-8), all on chromosome 4q25. Of these, 80 variants clustered upstream of PITX2, and two were located in LINC01438. Fine mapping identified two independent intergenic signals, with rs2200732 as the lead single-nucleotide polymorphism. The best PRS-only model achieved an AUC of >0.70, which was improved up to 0.737 in additive models incorporating both PRS and clinical variables. SHAP analysis consistently ranked PRS as the most influential predictor among the clinical variables included in this study. These results suggest that genetic risk, particularly at the established 4q25/PITX2 locus, contributes substantially to PAF susceptibility in this Japanese cohort and that PRS may improve early risk stratification when integrated with clinical risk factors.