Identification of asthma-related trans-acting epistatic eQTLs using Model-Based Multifactor Dimensionality Reduction (MB-MDR)

Epistasis is likely to underlie most complex traits, including gene expression, yet it is very difficult to detect using standard approaches. SNPs located inside a gene coding region or in its vicinity (i.e. ≤2 Mb from each 5’ and 3’ side) can influence the corresponding gene expression levels. These expression quantitative trait loci (eQTLs) are referred to as cisSNPs. In contrast, eQTLs that are outside the aforementioned gene range can also influence the gene’s expression, in which case, they are called transSNPS to that gene. In this study we considered significant cisSNPs previously identified via generalized least squares (GLS) regression modeling. We then identified those genes transcripts whose expression is regulated by cis/transSNP interaction. In this work we aimed at identifying transcripts whose expression is regulated by a cis/transSNP interactions using Model-Based Multifactor Dimensionality Reduction (MB-MDR) [2]. This model-free approach to detect trans-epistasis involves reducing a high-dimensional GxG space to GxG factor levels that either exhibit high evidence, low evidence or no evidence at all for their association to gene expression levels of interest.
Our protocol was applied on real-life data from the childhood asthma management program (CAMP) [1]. It involved coupling a traditional a priori eQTL search to an a posteriori trans-epistasis analysis to identify genetic modifiers to statistically significant cisSNPs. Such an approach allows to reveal previously unreported inter-dependencies that may be important in understanding of biological mechanisms underlying human complex diseases such as asthma.
The proposed protocol identified a large number trans-epistasis gene-gene effects of eQTLs.