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

Area of Science:

• Genetics
• Statistical Genetics
• Systems Biology

Background:

• Epistasis, or gene-gene interaction, is crucial in biology but challenging to model for complex traits due to polygenic interactions.
• Current models often fail to capture the complexity of polygenic epistasis effectively.
• A need exists for advanced statistical models to understand the genetic architecture of complex traits.

Purpose of the Study:

• To develop and validate a novel statistical model, Epistasis Factor Analysis (EFA), for modeling polygenic epistasis in complex traits.
• To improve the accuracy of polygenic prediction and increase the power to detect epistasis.
• To biologically interpret genetic effects by decomposing them into more homogeneous units.

Main Methods:

• Developed Epistasis Factor Analysis (EFA), a model that factorizes polygenic epistasis into interactions among latent epistasis factors (EFs).
• Mathematically characterized EFA and validated its performance using simulations against existing epistasis models.
• Applied EFA to predict yeast growth rates and analyze complex traits in the UK Biobank dataset.

Main Results:

• Simulations demonstrated that EFA outperforms current epistasis models when its assumptions are met.
• EFA significantly improved prediction accuracy for yeast growth rates, outperforming both additive and standard epistasis models.
• Analysis of UK Biobank data revealed statistically significant epistasis for four complex traits, with inferred EFs partially recovering known biological pathways.

Conclusions:

• Epistasis Factor Analysis (EFA) provides a more realistic and powerful approach to modeling complex trait epistasis.
• The findings suggest that epistasis plays a significant role in complex traits and can be biologically interpreted.
• EFA holds promise for advancing precision medicine and elucidating the genetic basis of diseases through GWAS results.