Semiparametric Bayesian variable selection for gene-environment interactions

Area of Science:

• Genetics and Bioinformatics
• Statistical Modeling
• Computational Biology

Background:

• Complex diseases arise from interactions between genetic variants and environmental factors.
• Studying gene-environment (G×E) interactions is crucial for understanding disease etiology.
• Existing Bayesian methods face challenges with high-dimensional data and complex environmental influences in G×E studies.

Purpose of the Study:

• To develop a novel semiparametric Bayesian variable selection model for G×E interactions.
• To simultaneously investigate linear and nonlinear G×E interactions.
• To enable structural identification distinguishing nonlinear interactions from main effects within a Bayesian framework.

Main Methods:

• Proposed a novel semiparametric Bayesian variable selection model.
• Incorporated spike-and-slab priors at individual and group levels for sparse effect identification.
• Developed a method for distinguishing nonlinear interactions from main effects.

Main Results:

• The proposed Bayesian method efficiently conducts variable selection for G×E interactions.
• Simulations demonstrate superior performance in identification and prediction compared to existing methods.
• The model successfully identifies significant main and interaction effects in high-dimensional SNP data.

Conclusions:

• The novel Bayesian approach offers a powerful tool for analyzing complex G×E interactions.
• This method has significant implications for high-throughput genetic profiling and disease research.
• It enhances the ability to elucidate disease etiology by accurately identifying genetic and environmental contributions.