SNPs were retained for analysis only if imputed successfully on both platforms, had a MAF > 0.01, and had high concordance in best-guess genotypes for the 327 individuals imputed twice (from Affymetrix and Illumina genotypes; the family-based association analysis requires use of best-guess genotypes). Specifically, high concordance was measured as discordance/heterozygosity > 0.04, where discordance is the proportion of individuals with discordant genotypes and heterozygosity = 2*MAF*(1-MAF), and MAF is the minor allele frequency estimated from the Illumina imputed set. The correction for heterozygosity removes the dependence of discordance rate on MAF. In total, 1,252,387 SNPs were available for association analyses, representing SNPs that are strongly validated for all samples. After QC, if only one individual from a monozygotic twin pair had been genotyped, the non-genotyped co-twin was assigned that genotype as well. The final genotyped sample in this study included 5117 individuals from 2567 families, including 797 monozygotic (MZ) twin pairs (1702 MZ twin individuals), of which 680 MZ individuals were assigned their cotwin’s genotype.
