We conducted the GWASs in 23andMe and UK-Biobank samples separately. Associations between the binary phenotype and SNPs were tested using a logistic regression model accounting for the effects of sex, age, ancestry, and genotype batch (and age2 in the UK-Biobank sample). The GWAS for UK-Biobank was performed in PLINK 1.951 and for 23andMe using an internally developed pipeline. We then meta-analysed the GWAS results from ICC, 23andMe, and UK-Biobank. Prior to conducting the meta-analysis, additional quality control of the summary statistics of each study was conducted in EasyQC52. Because of varying GWAS methods and sample characteristics (Supplementary Table S11), slightly different quality control criteria were used for the 3 samples (Supplementary Table S12). All 3 samples were aligned with the Haplotype Reference Consortium panel using the EasyQC R-package52, to ensure that rs-numbers and chromosome-basepair positions referred to the exact same variants and to correct for strand effects. Variants were deleted if they had a minor allele frequency (MAF) diverging more than 0.15 from that in the reference panel.
