A tree-based gene-environment interaction analysis with rare features

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Gene-environment (G-E) interaction analysis is crucial for understanding complex diseases.
• Joint G-E analysis faces challenges with high dimensionality, weak signals, and variable selection hierarchies, especially for rare genetic features.
• Existing methods for rare features are primarily for marginal analysis and not directly applicable to joint G-E interaction analysis.

Purpose of the Study:

• To develop a new gene-environment interaction analysis approach specifically tailored for rare genetic features.
• To address the limitations of existing methods in joint G-E interaction analysis concerning rare variants.
• To improve the identification of significant main effects and interactions involving rare genetic factors.

Main Methods:

• Developed a novel approach building upon a recent tree-based data aggregation technique for main-effect-only analysis.
• The method incorporates efficient information borrowing from neighboring rare features.
• Employs penalization for variable selection, regularized estimation, and respects the variable selection hierarchy.

Main Results:

• Simulation studies demonstrated more accurate identification of important interactions and main effects compared to competing methods.
• The proposed approach showed satisfactory prediction and stability performance in the NFBC1966 study analysis.
• Findings from the NFBC1966 study differed from those obtained using alternative methods.

Conclusions:

• The developed approach offers an effective strategy for joint gene-environment interaction analysis involving rare genetic features.
• It enhances the ability to detect complex disease associations by leveraging information from rare variants.
• The method provides a valuable tool for genetic epidemiology research, offering improved accuracy and performance.