Robust Gene-Gene Interaction Analysis in Genome Wide Association Studies

Area of Science:

• Genetics
• Statistical Genomics
• Bioinformatics

Background:

• Genome-wide association studies (GWAS) identify genetic variants linked to complex traits, but often explain limited heritability.
• Missing heritability is attributed to gene-gene (GG) and gene-environment (GE) interactions, and structural variants.
• Generalized multifactor dimensionality reduction (GMDR) detects interactions but is sensitive to outlying quantitative traits.

Purpose of the Study:

• To develop a robust GMDR estimation method to mitigate the impact of outlying traits on detecting gene interactions.
• To improve the accuracy and reliability of interaction detection in genetic association studies.

Main Methods:

• Proposed a robust GMDR estimation method using L-estimator and M-estimator techniques.
• Conducted simulation studies to compare robust GMDR with original MDR.
• Applied robust GMDR to a Korean population GWA dataset (8,577 samples) using HOMA-IR as the phenotype.

Main Results:

• Simulation studies demonstrated that robust GMDR outperforms the original MDR in detecting interactions.
• Robust GMDR successfully identified significant interaction effects involving the KCNH1 gene.
• The identified interactions were associated with the regulation of insulin secretion.

Conclusions:

• The proposed robust GMDR method effectively reduces the influence of outlying traits in interaction analysis.
• This method enhances the power to detect GG and GE interactions, contributing to understanding complex traits.
• Robust GMDR provides a valuable tool for genetic association studies, exemplified by its application to insulin resistance in the Korean population.