Cross-Ancestry Polygenic Prediction of Bipolar Disorder and Obsessive-Compulsive Disorder in India.

Abstract

Most genetic studies of psychiatric disorders have been conducted predominantly in populations of European ancestry, limiting evidence for the transferability of polygenic scores (PGS) to other populations. This cross-sectional study evaluated the predictive performance of European-derived PGS for bipolar disorder (BD) and obsessive-compulsive disorder (OCD) in an Indian sample. Participants (N = 2531) included individuals with BD (N = 463), OCD (N = 554), and controls (N = 1514) recruited from hospital- and community-based samples. Genotyping was performed using the Illumina Global Screening Array. PGS were derived from the summary statistics of European-ancestry genome-wide association studies using two Bayesian methods, PRS-CS-auto and SBayesRC. Principal component analyses demonstrated genetic overlap of the study population with other South Asian populations. BD PGS showed significant predictive performance, explaining approximately 6%-7% of liability-scale variance with PGS-only AUCs of ~0.71. OCD PGS showed lower but significant predictive performance, explaining ~2%-3% of variance with AUCs of approximately 0.59. Bayesian approaches showed better predictive performance than conventional clumping-and-thresholding. These findings demonstrate that European-derived psychiatric PGS can retain predictive information in an Indian population, although performance differed between disorders. The application of Bayesian PGS methods in this Indian cohort supports their utility for cross-ancestry prediction, while reinforcing the need for larger genomic studies in South Asian and other underrepresented populations.