Two-stage testing procedures with independent filtering for genome-wide gene-environment interaction

Area of Science:

• Genetics
• Biostatistics
• Epidemiology

Background:

• Genome-wide association studies (GWAS) often require methods to detect gene-environment interactions.
• Existing two-stage multiple testing procedures have limitations in validity and power.

Purpose of the Study:

• To elucidate general conditions for the validity and power of two-stage multiple testing procedures in GWAS.
• To propose extensions of these procedures using case-only estimators for gene-treatment interaction in randomized clinical trials (RCTs).

Main Methods:

• Developed a unified estimating equation approach to prove asymptotic independence between filtering and interaction test statistics.
• Applied methods to generalized linear models with canonical links, covering marginal association and interaction.
• Assessed performance through simulations and analysis of Women's Health Initiative clinical trial genetic data.

Main Results:

• Established general conditions for the validity and power of two-stage interaction testing procedures.
• Demonstrated the utility of a unified estimating equation approach for proving statistic independence.
• Validated proposed extensions in both simulated and real-world clinical trial data.

Conclusions:

• The study provides a theoretical framework and practical extensions for robust gene-environment interaction detection in GWAS and RCTs.
• The proposed methods enhance the reliability and power of statistical tests for genetic associations.
• Findings contribute to improved understanding of gene-environment and gene-treatment interactions in large-scale genetic studies.