Methylation quantitative trait loci analysis

To examine the association between methylation and SNPs within the 40 genes (with 50 kb flanking regions; gene boundaries are available in Additional file 13: Table S10), we used R package “MatrixEQTL” [81]. Only polymorphic SNPs (N = 46,780) were tested using linear model setting with a cis distance of 2.5 Mb. The longest gene was approximately 2.4 Mb (Additional file 13: Table S10); thus, a cis distance of 2.5 Mb was chosen to ensure that all possible combinations of SNPs, and the CpG within a gene are tested. All association tests between remaining SNP and CpGs were considered trans. Genotypes were coded as copies of effect allele (0, 1, or 2) and methylation data was extracted for the 55 probes from the CPACOR normalized data set used in the EWAS. We included, age, sex, BMI, white blood cell counts, control probe PCs, and cotinine as covariates in the model. In addition, owing to the relatedness in our sample, we included a covariance matrix based on CpG sites being tested in the model. Results of meQTL analysis were visualized using R package “RCircos” [82].