A Bayesian approach to Mendelian randomization with multiple pleiotropic variants.

Area of Science:

• Genetics
• Statistical Genetics
• Epidemiology

Background:

• Mendelian randomization (MR) is a powerful tool for inferring causality.
• Pleiotropy, where genetic variants affect the outcome through pathways other than the exposure, is a major challenge in MR.
• Existing MR methods may be sensitive to violations of core assumptions.

Purpose of the Study:

• To develop a flexible Bayesian framework for Mendelian randomization that explicitly models and accounts for pleiotropic effects.
• To provide robust causal inference in the presence of potential instrument bias and correlated instruments.
• To extend the Bayesian MR framework to handle multiple exposures and estimate direct/indirect effects.

Main Methods:

• A Bayesian statistical model is proposed for Mendelian randomization.
• Pleiotropic effects are modeled as unknown parameters with a shrinkage prior distribution, assuming a subset are zero.
• Inference is performed using Markov chain Monte Carlo (MCMC) methods to obtain posterior distributions for causal effects.

Main Results:

• The proposed Bayesian MR method demonstrates robustness to directional pleiotropy and moderate instrument correlation.
• Simulation experiments validate the performance of the method.
• The framework can be extended to analyze multiple exposures and disentangle direct and indirect causal effects.

Conclusions:

• The Bayesian approach offers a flexible and robust alternative for Mendelian randomization analysis, particularly when pleiotropy is a concern.
• This framework facilitates the incorporation of external information and handling of multiple uncertainties in causal inference.
• The developed model provides a foundation for future advancements in Bayesian Mendelian randomization.