Fig. 1. Overview of this study.

a Japanese individuals from seven major regions of Japan were genotyped. Phased and imputed genotypes were linkage disequilibrium (LD)-pruned and formatted as an input for dimensionality reduction methods. b We applied five dimensionality reduction methods to the genotype. We further applied PCA–UMAP to the subpopulations in an attempt to identify even finer substructures. c We performed fineSTRUCTURE, ADMIXTURE, and phylogenetic analyses in each subpopulation identified in b. d We investigated how the identified subpopulations affected polygenic risk predictions in a phenome-wide scale.