In order to make PGIs for the UK Biobank (UKB) without having to exclude the entire UKB from the discovery GWAS, we split the UK Biobank sample into three equal-sized partitions and, for each partition, used the summary statistics from the other two partitions when generating its PGI. The first partition (UKB1) is composed of UKB participants with brain-scan data (as indicated by data field 12188), all pairs of UKB participants related up to second degree, and the pairs of relatives of third-degree relatedness with greatest relatedness. Pairs of individuals of third-degree relatedness were ordered based on the maximum relatedness coefficient they have with another participant and assigned to the first partition in decreasing relatedness order until the partition was full. Remaining individuals with third-degree relatives were assigned to the second partition. Finally, individuals with no third degree or closer relatives were randomly assigned to the second (UKB2) or third (UKB3) partition.
