Joint Analysis of Multiple Interaction Parameters in Genetic Association Studies

Area of Science:

• Human genetics and epidemiology
• Statistical genetics
• Complex disease research

Background:

• Growing human genetic and epidemiologic data fuels interest in gene-environment (G-E) interactions.
• Challenges exist in jointly testing numerous interactions between multiple single nucleotide polymorphisms (SNPs) and exposures.

Purpose of the Study:

• To compare the performance of four fixed-effect joint analysis approaches for G-E interactions.
• To evaluate methods using simulated data and large human cohort datasets.
• To provide guidelines for analyzing G-E interactions in large human cohorts.

Main Methods:

• Compared four joint analysis approaches: omnibus test, multi-exposure genetic risk score (GRS) test, multi-SNP environmental risk score (ERS) test, and GRS-ERS test.
• Utilized simulated data with up to 10 exposures and 300 SNPs, exploring linear and logistic regression with Wald, Score, and likelihood ratio tests (LRT).
• Applied methods to human cohort data (n=37,664) for type 2 diabetes (T2D), obesity, hypertension, and coronary heart disease.

Main Results:

• GRS-based approaches demonstrated superior robustness and power, particularly when G-E effects aligned with marginal genetic and environmental effects.
• Severe miscalibration of joint statistics was observed in logistic models with low events-per-variable using Wald or LRT.
• The GRS-ERS approach identified nominally significant G-E interactions for T2D, obesity, and hypertension in real data.

Conclusions:

• GRS-based methods, especially GRS-ERS, are recommended for analyzing multiple G-E interactions in large human cohorts.
• Caution is advised regarding joint statistic calibration in logistic models with sparse data.
• This study offers practical guidance for G-E interaction analysis in contemporary human genetic and environmental datasets.