All PGIs in the initial release of the Repository were constructed in Plink279 using imputed genotype probabilities. Prior to constructing the PGIs, we adjusted the SNP weights for linkage disequilibrium (LD) using LDpred30. We estimated the LD patterns using genotype data from the public release of the HRC Reference Panel (version 1.1) after applying the following quality-control filters. First, we limited the set of variants to HapMap3 SNPs and filtered out variants with genotyping call rate <0.98 and individuals with genotype missingness rate >0.02. Next, we calculated the genomic relatedness matrix and dropped one individual out of each pair with relatedness coefficient >0.025. We clustered the remaining individuals based on their identity-by-state distances using Plink1.9 and dropped an individual if the Z-score corresponding to their distance to their nearest neighbour is less than −5. In the remaining sample that we fed into LDpred for LD estimation, there were 1,214,408 SNPs and 14,028 individuals. At the coordination step of LDpred, we used the option “--max-freq-discrep” in order to exclude markers that have a frequency discrepancy greater than 0.1 between the summary
