For the purposes of pairing genes and SNPs in order to link the gene expression study to the GWAS, we selected the proximal SNP (within 150 kb of the gene) with the best evidence of being a cis eQTL. We used the R package, ‘ICE’ (82), which maps eQTLs using a linear mixed-effects model framework. ICE was designed to take into account covariance across arrays that can lead to spurious trans-eQTL results. By replacing the array covariance matrix with the additive covariance matrix for the Hutterite pedigree, we were able to control for the additive polygenic effect of interrelatedness. Variance components were estimated using a restricted maximum likelihood method, and a P-value for each SNP was calculated using an F-test. The proximal SNP with the smallest P-value was then classified as the ‘eQTL’ for each gene.
