Robust gene-environment interaction analysis using penalized trimmed regression.

Area of Science:

• Biostatistics
• Genetics
• Epidemiology

Background:

• Gene-environment (G-E) interactions play a critical role in the development and progression of complex diseases.
• Current methods for identifying G-E interactions often lack robustness against data outliers and contamination.
• Addressing these limitations is crucial for accurate disease etiology research.

Purpose of the Study:

• To develop a novel, robust approach for identifying gene-environment interactions.
• To enhance the reliability of G-E interaction analysis in biomedical and epidemiological studies.
• To improve the accuracy and stability of identifying key G-E interactions in complex disease research.

Main Methods:

• Utilized trimmed regression techniques within a joint modeling framework.
• Employed a robust data-driven criterion and stability selection to identify outlier-free subsets.
• Developed an effective penalization strategy to detect G-E interactions while preserving hierarchical structures (main effects and interactions).

Main Results:

• The proposed robust approach demonstrated superior performance compared to existing methods in extensive simulations.
• The method achieved better prediction accuracy and stability in identifying G-E interactions.
• Significant findings were observed in analyses of TCGA cutaneous melanoma and breast invasive carcinoma data.

Conclusions:

• The novel robust G-E interaction identification method offers improved accuracy and reliability.
• This approach effectively addresses limitations of existing methods concerning data outliers.
• The findings have implications for understanding complex diseases and advancing precision medicine through G-E interaction analysis.