Estimating and testing direct genetic effects in directed acyclic graphs using estimating equations.

Area of Science:

• Genetics
• Biostatistics
• Causal Inference

Background:

• Distinguishing direct and indirect genetic effects is crucial for building functional genetic models.
• Existing methods may not adequately account for complex relationships between genetic variants, phenotypes, and confounders.

Purpose of the Study:

• To propose and evaluate a novel method, Causal Inference based on Estimating Equations (CIEE), for valid statistical inference of direct genetic effects.
• To assess CIEE's performance against traditional methods in simulations and a real-world genetic dataset.

Main Methods:

• Utilized a directed acyclic graph (DAG) framework incorporating genetic variants, primary and intermediate phenotypes, and confounders.
• Developed the Causal Inference based on Estimating Equations (CIEE) method using robust Huber-White sandwich standard errors.
• Compared CIEE with multiple regression, structural equation modeling, and sequential G-estimation for quantitative and time-to-event traits.

Main Results:

• CIEE demonstrated valid estimation and inference, effectively isolating direct genetic effects by removing intermediate phenotype influences.
• CIEE proved robust against measured and unmeasured confounding factors affecting indirect genetic effects.
• Simulation studies revealed inflated type I errors in other methods, while CIEE maintained statistical validity.
• Analysis of the Genetic Analysis Workshop 19 dataset identified genetic variants missed by traditional regression analyses.

Conclusions:

• CIEE offers a robust and accurate approach for estimating direct genetic effects in complex genetic association studies.
• The method is computationally efficient, broadly applicable, and available as an R package, facilitating wider adoption.