A genetic algorithm for detecting higher-order gene-by-environment interactions using case-parent triads

Abstract

We developed a genetic algorithm, called Exposure-related Genetic Algorithm for Detecting Genetic Epistasis using Triads and Siblings (E-GADGETS), to search for risk-associated interactions among multiple single-nucleotide polymorphisms (SNPs) and non-genetic “exposures” (G×G×E interactions) using case-parent triads. In E-GADGETS, a non-genetic “exposure” may comprise factors ranging from health-related behaviors and environmental agents to disease subcategories. We demonstrate through simulation that, when searching 10 000 or 50 000 candidate SNPs, E-GADGETS identifies G×G×E interactions involving three or four SNPs and categorical or continuous exposures. We also describe how to adapt existing triad-based epistasis-mining algorithms to search for G×G×E effects provided that E is binary and the set of candidate SNPs is small enough, but E-GADGETS outperforms those methods. We further describe permutation-based inferential procedures and a visualization approach for results. When applied to candidate SNPs from publicly available case-parent data, E-GADGETS detected evidence of genetic interactions with maternal exposure to environmental tobacco smoke associated with cleft-palate-only in Asians.