Causal mediation analysis for time-varying heritable risk factors with Mendelian randomization

Area of Science:

• Genetics
• Epidemiology
• Computational Biology

Background:

• Causal inference is vital in clinical research, often using Mendelian Randomization (MR) to overcome confounding when randomized trials are infeasible.
• Traditional MR assumes static risk factors, failing to capture dynamic effects crucial for understanding disease development over a lifetime.
• Existing life-course MR methods face limitations due to small GWAS cohort sizes and challenges with correlated longitudinal data.

Purpose of the Study:

• To introduce FLOW-MR, a computational framework for estimating causal structural equations from summary statistics of genome-wide association studies (GWAS).
• To enable robust inference of direct, indirect, and path-wise causal effects for temporally ordered traits.
• To address limitations in current life-course MR, particularly with polygenic traits and weak genetic instruments.

Main Methods:

• FLOW-MR utilizes GWAS summary statistics to model causal relationships between traits measured at different life stages.
• The approach incorporates a spike-and-slab prior to handle extreme polygenicity and weak instrument challenges.
• It estimates causal structural equations, allowing for nuanced analysis of dynamic risk factor effects.

Main Results:

• FLOW-MR demonstrated superior efficiency and reliability compared to existing methods, especially with noisy data.
• A childhood-specific protective effect of Body Mass Index (BMI) on breast cancer risk was identified.
• The evolving causal impacts of BMI, systolic blood pressure, and cholesterol on stroke risk were analyzed over time.

Conclusions:

• FLOW-MR provides a powerful computational tool for life-course causal inference using readily available GWAS summary statistics.
• The method enhances understanding of dynamic risk factor trajectories and their complex relationships with disease.
• Findings highlight the importance of considering temporal dynamics in causal analyses for improved clinical and public health insights.