Finding the Epistasis Needles in the Genome-Wide Haystack

Abstract

Genome-wide association studies (GWAS) have dominated the field of human genetics for the past 10 years. This study design allows for an unbiased, dense exploration of the genome and provides researchers with a vast array of SNPs to look for association with their trait or disease of interest. GWAS has been referred to as finding needles in a haystack and while many of these "needles," or SNPs associating with disease, have been identified, there is still a great deal of heritability yet to be explained. The missing or phantom heritability is due, at least in part, to epistasis or gene-gene interactions, which have not been extensively explored in GWAS. Part of the challenge for epistasis analysis in GWAS is the sheer magnitude of the search and the computational complexity associated with it. An exhaustive search for epistasis models is not computationally feasible; thus, alternate approaches must be considered. In this chapter, these approaches will be reviewed briefly, and the incorporation of biological knowledge to guide this process will be further expanded upon. Real biological data examples where this approach has yielded successful identification of epistasis will also be provided. Epistasis has been known to be important since the early 1900s; however, its prevalence in mainstream research has been somewhat overshadowed by molecular technology advances. Due to the increasing evidence of epistasis in complex traits, it continues to emerge as a likely explanation for missing heritability.

Genetic dissection of femoral and tibial microarchitecture

“…While this interaction is nominally significant, we are not convinced that this interaction is genuine for two reasons: 1) our search for epistatic interactions does not correct for the many secondary tests we performed, and 2) neither loci in this epistatic pairing are associated with a significant additive effect for this or any of the other 49 traits (Chr 1 at 100.7 Mb and Chr 15 at 33.4 Mb). We therefore conclude that epistatic interactions have relatively modest effects that are not detectible with our modest sample size (71,72) .…”

Multilocus approaches for the measurement of selection on correlated genetic loci

“…; Wang et al . ; Ritchie , ). Our understanding of speciation would benefit from measures of selection that explicitly incorporate genotype–environment interactions or that tie selection to trait genetics.…”

“…Orr 1995;Turelli & Orr 2000;Gavrilets 2004;Orr 2005). Dominance can readily be incorporated into whole-genome regression models, such as BSLMMs, and the same is true in principle for epistasis but the number of genotype combinations present a daunting, but not insurmountable, computational challenge (Zhang & Liu 2007;Jiang et al 2009;Wang et al 2010;Ritchie , 2011Ritchie , , 2015. Our understanding of speciation would benefit from measures of selection that explicitly incorporate genotype-environment interactions or that tie selection to trait genetics.…”

Polygenic link between blood lipids and amyotrophic lateral sclerosis

“…So far, two major explanations have been proposed and further investigated: 1) the numerator h 2 SNP is underestimated, because current SNP-based methods haven't included the rare SNPs or other types of genetic variants (e.g. copy number variations, insertions and deletions), nor accounting for gene-environment interactions [35]; 2) the denominator h 2 Twin is overestimated in classical twin studies, due to potential violation of EEA, or falsely ascribing true nonadditive effects to A (and thereby the narrow-sense heritability is overestimated) [36][37][38].…”