For the eQTL analyses, we used cerebellum44 (n = 153; publicly available in GEO: GSE35974) and adipose45 (n=62; publicly available in GEO: GSE40234); details of the samples, the quality control procedure and the normalization approach used were previously described21. For adipose, the samples were chosen to be at the tails of the distribution of insulin sensitivity matched on age, gender and natural log of BMI. Because the adipose samples included individuals of African ancestry, we performed principal component (PC) analysis and tested each marker in the locus for the additive effect of allele dosage on the residual expression phenotype (IRX3 and FTO) after adjusting for the PCs (n=2). Similarly, for cerebellum, we performed linear regression on the residual expression phenotype (IRX3 and FTO) after adjusting for sex and pH to evaluate the additive effect of allele dosage at each marker in the locus; the non-diseased samples (obtained from the Stanley Medical Research Institute) included in the analysis were of European descent and thus the PCs were not used to generate the residual expression. Significance (p-value) was evaluated using the t-statistic from the regression.
