Using available summary statistics form published GWAS as training datasets, we calculated polygenic risk scores (PRS) for individuals in the iPSYCH2015 sample using LDpred2132. Summary statistics were filtered for an imputation info score INFO≥0.9 if available. When using external summary stats not processed using Ricopili or imputed using different imputation references, we excluded all ambiguous markers to avoid potential strand conflicts. To improve performance of the scores and avoid including artefacts from batch effects, we restricted the summary stats to include only SNPs known to be present in both the iPSYCH2012 and iPSYCH2015i data at a reasonable quality (info score 0.6 and MAF 0.05). This step also checked for allele flips.
