Using linear regression, in each cohort we estimated how much variance in the sex- and age-adjusted BMI and height phenotypes (adjusted R2) was explained by the methylation- and genetic-profile scores, both individually and combined. We also looked for any evidence of interaction between the methylation- and genetic-profile scores. For each trait in each cohort, we ran the following four regression models and extracted the proportion of variance explained from each:Model 1: trait ∼ MWAS score
Model 2: trait ∼ GWAS score
Model 3: trait ∼ MWAS score + GWAS score
Model 4: trait ∼ MWAS score + GWAS score + (MWAS score × GWAS score)