Integrating Multi-Omics Data for Gene-Environment Interactions.

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Gene-environment (G×E) interactions are crucial for complex disease etiology, extending beyond individual genetic or environmental effects.
• Existing variable selection methods struggle to integrate multidimensional omics data and capture structured sparsity for G×E interaction analysis.

Purpose of the Study:

• To develop a novel penalized variable selection method for integrating multi-omics data in G×E interaction studies.
• To address the limitations of current methods in handling structured sparsity and multidimensional omics measurements.

Main Methods:

• Developed an integrative model incorporating sparse dimensionality reduction for gene expression regulators.
• Accommodated a sparse bi-level structure to link disease outcomes with multiple effects in G×E studies.
• Utilized penalized variable selection for dissecting G×E interactions within multi-omics data.

Main Results:

• Simulation studies demonstrated superior identification of G×E interactions and regulators compared to alternative methods.
• The integrative model achieved improved prediction performance in two high-dimensional, multi-omics G×E lung cancer studies.
• Identified biologically relevant findings with significant implications for lung cancer research.

Conclusions:

• The novel integrative model effectively dissects G×E interactions by leveraging multi-omics data and structured sparsity.
• This approach enhances the identification of disease-related factors and improves predictive accuracy in complex diseases.
• The method offers a promising tool for future G×E interaction research, particularly with high-dimensional omics datasets.