Mapping eQTL by leveraging multiple tissues and DNA methylation

DNA methylation is an important tissue-specific epigenetic event that influences transcriptional regulation of gene expression. Differentially methylated CpG sites may act as mediators between genetic variation and gene expression, and this relationship can be exploited while mapping multi-tissue expression quantitative trait loci (eQTL). Current multi-tissue eQTL mapping techniques are limited to only exploiting gene expression patterns across multiple tissues either in a joint tissue or tissue-by-tissue frameworks. We present a new statistical approach that enables us to model the effect of germ-line variation on tissue-specific gene expression in the presence of effects due to DNA methylation.

Our method efficiently models genetic and epigenetic variation to identify genomic regions of interest containing combinations of mRNA transcripts, CpG sites, and SNPs by jointly testing for genotypic effect and higher order interaction effects between genotype, methylation and tissues. We demonstrate using Monte Carlo simulations that our approach, in the presence of both genetic and DNA methylation effects, gives an improved performance (in terms of statistical power) to detect eQTLs over the current eQTL mapping approaches. When applied to an array-based dataset from 150 neuropathologically normal adult human brains, our method identifies eQTLs that were undetected using standard tissue-by-tissue or joint tissue eQTL mapping techniques. As an example, our method identifies eQTLs by leveraging methylated CpG sites in a LIM homeobox member gene (LHX9), which may have a role in the neural development.

Our score test-based approach does not need parameter estimation under the alternative hypothesis. As a result, our model parameters are estimated only once for each mRNA - CpG pair. Our model specifically studies the effects of non-coding regions of DNA (in this case, CpG sites) on mapping eQTLs. However, we can easily model micro-RNAs instead of CpG sites to study the effects of post-transcriptional events in mapping eQTL. Our model's flexible framework also allows us to investigate other genomic events such as alternative gene splicing by extending our model to include gene isoform-specific data.