# A method for finding epistatic effects of maternal and fetal variants

## TL;DR

This paper introduces an improved method to detect genetic interactions between maternal and fetal DNA that influence pregnancy outcomes and offspring diseases.

## Contribution

The paper extends the GADGETS algorithm to include maternal SNPs and outperforms existing methods in detecting maternal-fetal epistatic effects.

## Key findings

- GADGETS successfully detected simulated multi-locus effects involving 3-5 SNPs in simulations.
- GADGETS outperformed other epistasis-mining algorithms when applied to a large list of candidate SNPs.
- Applied to real data, GADGETS identified maternal-fetal SNP sets potentially linked to orofacial clefting.

## Abstract

Pregnancy involves a double genome, and genetic variants in the mother and her fetus can act together to influence risk for pregnancy complications, adverse pregnancy outcomes, and diseases in the offspring. Large search spaces have hindered the discovery of sets of single nucleotide polymorphisms (SNPs) that act epistatically.

Previously, we proposed a method for case-parent studies, called the Genetic Algorithm for Detecting Genetic Epistasis using Triads or Siblings (GADGETS), that can reveal autosomal epistatic SNP-sets in the child’s genome. Here we incorporate maternal SNPs, thereby extending GADGETS to nominate SNP-sets containing offspring loci only, maternal loci only, or both. We use a permutation procedure to impose a preference for epistatic over outcome-related but non-epistatic SNP sets. Our maternal-fetal extension uses case-complement-sibling pairs together with mother-father pairs, exploiting Mendelian transmission and a mating-symmetry assumption.

In simulations of 1,000 case-parents triads with 10,000 candidate SNPs, GADGETS successfully detected simulated multi-locus effects involving 3-5 SNPs but was somewhat less successful at distinguishing epistatic SNPs from sets of non-epistatic SNPs that each conferred high risk independently. Though the epistasis-mining algorithms MDR-PDT, TrioFS, and EPISFA-LD were originally designed to find epistatic offspring variants, we generalize them to include maternal SNPs and search more broadly. GADGETS outperformed those competitors and could successfully mine a much larger list of candidate SNPs. Applied to dbGaP data, GADGETS nominated several multi-SNP maternal-fetal sets as potentially-interacting risk factors for orofacial clefting.

The extended version of GADGETS can mine for epistasis that involves maternal SNPs.