Imputed genotypes for both samples were derived using the 1000 Genomes reference panels as previously reported (Edwards et al., 2017; Webb et al., 2017). Quality control procedures for genetic analyses of both the ALSPAC and S4S samples have been previously described (Fatemifar et al., 2013; Webb et al., 2017), and those within-sample approaches were applied for the current analyses (see Supplementary Information). Briefly, individual DNA samples and markers were excluded based on excess missingness (>5% for both samples), deviations from Hardy-Weinberg Equilibrium (p<5e-7 for ALSPAC, p<5e-6 for S4S), and minor allele frequency (<0.01 for ALSPAC, <0.005 for S4S). Cryptic relatedness was calculated using pi-hat and related individuals were excluded.
