A powerful and efficient two-stage method for detecting gene-to-gene interactions in GWAS

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Gene-gene interactions (epistasis) are implicated in the missing heritability of complex traits.
• Genome-wide epistasis detection is computationally challenging and suffers from low statistical power due to numerous tests.

Purpose of the Study:

• To propose a computationally feasible and statistically powerful two-stage method for genome-wide epistasis detection.
• To reduce the multiple testing burden in genome-wide association studies (GWAS) for epistasis.

Main Methods:

• A two-stage approach involving a pre-assessment of genotype independence in cases, followed by logistic regression score tests for promising locus pairs.
• The method filters candidate pairs based on non-equilibrium frequencies before performing computationally intensive score tests.

Main Results:

• The proposed method significantly reduces the multiple testing burden, making genome-wide epistasis studies feasible.
• Demonstrated strong control of type I error and favorable power, particularly under model misspecification.
• Identified potential SNP-SNP interactions in Parkinson's disease across multiple cohorts.

Conclusions:

• The novel two-stage method offers a practical solution for genome-wide epistasis detection in complex traits.
• This approach enhances the feasibility and power of GWAS for uncovering gene-gene interactions.
• The method has been successfully applied to Parkinson's disease data, revealing potential interactions.