First, the data of each participating cohort were analyzed individually (see Supplementary Note 2 for details). Genetic data were imputed using either the Haplotype Reference Consortium (HRC)94 or 1000 Genomes Project Phase 3 reference panels95. The resulting GWAS summary statistics were then harmonized before a conjoint meta-analysis of all autosomes was conducted. Each summary statistic dataset was transformed to the ‘daner’ file format following RICOPILI96 specifications. All variants had to meet the following criteria for inclusion: minor allele frequency (MAF) > 1% in cases and controls, INFO score > 0.8 and <1.2. If the effect measure, P value or s.e. was missing or was out of bounds (infinite), the SNP was removed. Once cleaned summary statistics were produced, all datasets were aligned to the HRC reference panel. If variants were reported on different strands, they were flipped to the orientation in the HRC reference. Furthermore, strand-ambiguous A/T and C/G SNPs were removed if their MAF was >0.4. In the case that A/T and C/G SNPs showed a MAF < 0.4, allele frequencies were compared to frequencies in the HRC reference.
