For all models, we built a PRS for each individual j in our test set (in all cases, there is no overlap between GWAS samples and PRS samples) using variant effect size estimates from GWAS as follows: (1)PRSj=∑iMAj,iβi, where M is the total number of SNPs shared between GWAS summary statistics and post-imputation genotype data of EAS individuals, i is the ith SNP in the model, Aj,i is the allelic dosage of the trait-increasing allele i in individual j and βi is the estimated effect size of allele i from GWAS. We calculated PRS using PLINK2.
