For our score derivation, we used summary statistics from recent GWAS studies conducted primarily among participants of European ancestry for five diseases16,30–33 and a linkage disequilibrium reference panel of 503 European samples from 1000 Genomes phase 3 version 5.34 UK Biobank samples were not included in any of the five discovery GWAS studies. DNA polymorphisms with ambiguous strand (A/T or C/G) were removed from the score derivation. For each disease, we computed a set of candidate genome-wide polygenic scores (GPS) using the LDPred algorithm and a pruning and threshold derivation strategies.
