Identification of gene-environment interactions with marginal penalization

Area of Science:

• Genetics and Bioinformatics
• Statistical Genomics
• Computational Biology

Background:

• Gene-environment (G-E) interaction analysis is crucial for understanding complex diseases.
• Current marginal analysis methods often fail to respect the main effects-interactions hierarchy and may perform poorly with regularization.
• Existing solutions for these limitations are often overly complex.

Purpose of the Study:

• To propose a novel marginal penalization approach for G-E interaction analysis.
• To address limitations of standard methods, including hierarchy violations and poor regularization performance.
• To develop an intuitive and effective framework for G-E interaction analysis.

Main Methods:

• Developed a marginal penalization approach with a novel penalty function.
• Designed a framework aligned with recent joint analysis methods.
• Evaluated the approach through simulations and real-world data analysis (SNP and gene expression data).

Main Results:

• The proposed marginal penalization approach outperforms popular significance-based and simple penalization methods in simulations.
• The method effectively handles the main effects-interactions hierarchy.
• Demonstrated promising findings in the analysis of single-nucleotide polymorphism and gene expression data.

Conclusions:

• The novel marginal penalization approach offers a more coherent and intuitive framework for G-E interaction analysis.
• This method provides a superior alternative to existing techniques, particularly when regularization is needed or hierarchy must be preserved.
• The approach shows potential for advancing the understanding of complex diseases through G-E interaction studies.