Variable selection for single-index varying-coefficients models with applications to synergistic G × E interactions.

Area of Science:

• Environmental Epidemiology
• Statistical Genetics
• Bioinformatics

Background:

• Simultaneous exposure to multiple environmental risk factors (Es) can elevate disease risk beyond individual effects.
• Synergistic gene-environment interactions (synG × E) describe how combined genetic and multiple environmental exposures impact disease risk.
• Single-index varying-coefficients models (SIVCM) are used to analyze synG × E effects.

Purpose of the Study:

• To propose a unified variable selection approach for SIVCM to model synG × E.
• To estimate and select important genetic and environmental variables contributing to synergistic interactions.
• To identify nonlinear synG × E, no synG × E, and no genetic effects.

Main Methods:

• Developed a unified variable selection approach for SIVCM.
• The method estimates varying, non-zero constant, and zero effects for gene variables.
• Selected important environmental variables involved in synergistic interactions.
• Theoretically evaluated the oracle property of the variable selection approach.

Main Results:

• The proposed method effectively estimates and selects important variables for synG × E.
• Simulation studies demonstrated strong finite sample performance for both continuous and discrete gene variables.
• Application to a real dataset confirmed the method's practical utility.

Conclusions:

• The developed approach provides a robust tool for identifying synergistic gene-environment interactions.
• This method has broad applicability in research aiming to understand complex disease etiology.
• It enables a deeper understanding of how multiple environmental exposures jointly influence genetic risks.