Factorizing polygenic epistasis improves prediction and uncovers biological pathways in complex traits

Abstract

AbstractEpistasis is central in many domains of biology, but it has not yet proven useful for complex traits. This is partly because complex trait epistasis involves polygenic interactions that are poorly captured in current models. To address this gap, we develop a new model called Epistasis Factor Analysis (EFA). EFA assumes that polygenic epistasis can be factorized into interactions between a few latent pathways, or Epistasis Factors (EFs). We mathematically characterize EFA and use simulations to show that EFA outperforms current epistasis models when its assumptions approximately hold. Applied to predicting yeast growth traits, EFA outperforms the additive model for several traits with large epistasis heritability and uniformly outperforms the standard epistasis model, which we replicate in a second dataset. We also apply EFA to four traits in the UK Biobank and find statistically significant evidence for epistasis in male testosterone. Moreover, we find that the inferred EFs partly recover known biological pathways. Our results demonstrate that realistic statistical models can identify meaningful epistasis in complex traits, indicating that epistatic models have promise for precision medicine and characterizing the biology underlying GWAS results.