Genome-wide association studies across diverse populations may help validate and confirm genetic contributions to risk of disease. We estimated the extent of population stratification as well as the predictive accuracy of polygenic scores (PGS) derived from European samples to a data set from India. We analysed 2685 samples from two data sets, a population neurodevelopmental study (cVEDA) and a hospital-based sample of bipolar affective disorder (BD) and obsessive-compulsive disorder (OCD). Genotyping was conducted using Illumina's Global Screening Array. Population structure was examined with principal component analysis (PCA), uniform manifold approximation and projection (UMAP), support vector machine (SVM) ancestry predictions, and admixture analysis. PGS were calculated from the largest available European discovery GWAS summary statistics for BD, OCD, and externalizing traits using two Bayesian methods that incorporate local linkage disequilibrium structures (PGS-CS-auto) and functional genomic annotations (SBayesRC). Our analyses reveal global and continental PCA overlap with other South Asian populations. Admixture analysis revealed a north-south genetic axis within India (FST 1.6%). The UMAP partially reconstructed the contours of the Indian subcontinent. The Bayesian PGS analyses indicates moderate-to-high predictive power for BD. This was despite the cross-ancestry bias of the discovery GWAS dataset, with the currently available data. However, accuracy for OCD and externalizing traits was much lower. The predictive accuracy was perhaps influenced by the sample size of the discovery GWAS and phenotypic heterogeneity across the syndromes and traits studied. Our study results highlight the accuracy and generalizability of newer PGS models across ancestries. Further research, across diverse populations, would help understand causal mechanisms that contribute to psychiatric syndromes and traits.