Abstract DNA methylation (DNAme) is the most commonly studied epigenetic mark in human populations. DNAme has gained attention in the Developmental Origins of Health and Disease field due to its gene expression regulation and potential long-term stability. Genetic variation and environmental exposures are amongst the main factors influencing inter-individual DNAme variability. However, the proportion and genomic distribution of their individual, additive and interactive effects on the DNA methylome remains unclear. Here, we introduce RAMEN, a Findable, Accessible, Interoperable, and Reusable (FAIR) framework tailored for DNAme microarrays. Using machine learning and statistical techniques, RAMEN models and dissects gene-environment contributions to genome-wide Variably Methylated Regions (VMRs), while controlling for spurious associations. To comprehensively test the power of RAMEN, we analyzed and characterized VMRs from cord blood samples from two independent cohorts (CHILD and PREDO; overall n=1,662). We identified genetics as a consistent key contributor to DNAme variability, usually in additive and interactive combinations with the environment, with genetic terms explaining the largest proportion of DNAme variance, compared to environmental and interaction terms. Operationalizing RAMEN as an R package to conduct scalable genome-exposome contribution analyses, our results highlighted the importance of genetic variation in sculpting DNAme patterns in early life.
