Gene-methylation interactions: Discovering region-wise DNA methylation levels that modify SNP-associated disease risk

Abstract

The genetic code is tightly linked to epigenetic instructions as to what genes to express, and when and where to express them. The most studied epigenetic mark is DNA methylation at CpG dinucleotides. Todays technology enables a rapid assessment of DNA sequence and methylation levels at a single-site resolution for hundreds of thousands of sites in the human genome, in thousands of individuals at a time. Recent years have seen a rapid increase in epigenome-wide association studies (EWAS) searching for the causes of risk for genetic diseases that previous genome-wide association studies (GWAS) could not pinpoint. However, those single-omics data analyses led to even more questions and it has become clear that only by integrating data one can get closer to answers. Here, we propose two new methods within genetic association analyses that treat the level of DNA methylation at a given CpG site as environmental exposure. Our analyses search for statistical interactions between a given allele and DNA methylation (GxMe), and between a parent-of-origin effect and DNA methylation (PoOx Me). The new methods were implemented in the R package Haplin and were tested on a dataset comprising genotype data from mother-father-child triadsm with DNA methylation data from the children only. The phenotype here was orofacial clefts (OFC), a relatively common birth defect in humans, which is known to have a genetic origin and an environmental component possibly mediated by DNA methylation. We found no significant PoOxMe interactions and a few significant GxMe interactions. Our results show that the significance of these interaction effects depends on the genomic region in which the CpGs reside and on the number of strata of methylation level. We demonstrate that, by including the methylation level around the SNP in the analyses, the estimated relative risk of OFC can change significantly. We also discuss the importance of including control data in such analyses. The new methods will be of value for all the researchers who want to explore genome- and epigenome-wide datasets in an integrative manner. Moreover, thanks to the implementation in a popular R package, the methods are easily accessible and enable fast scans of the genome- and epigenome-wide datasets.