Disentangling the effects of traits with shared clustered genetic predictors using multivariable Mendelian...

Area of Science:

• Genetics
• Epidemiology
• Biostatistics

Background:

• Disentangling causal pathways from linked genetic variants to disease outcomes is challenging.
• Genetic variants in the chemokine receptor gene cluster are associated with cytokines and stroke risk.
• Standard Mendelian randomization methods struggle with highly correlated genetic variants in clusters.

Purpose of the Study:

• To develop and validate multivariable Mendelian randomization methods for analyzing correlated genetic variants in clusters.
• To address the Goldilocks dilemma of variant selection in clustered genetic regions.
• To identify the specific causal risk factor for stroke within the chemokine gene cluster.

Main Methods:

• Proposed multivariable Mendelian randomization methods utilizing principal component analysis.
• Reduced correlated genetic variants into orthogonal components for instrumental variables.
• Validated methods through simulations assessing precision and numerical stability.

Main Results:

• Simulations demonstrated improved precision and reduced sensitivity to numerical instability.
• The proposed methods effectively handle highly correlated genetic variants.
• Identified monocyte chemoattractant protein-1 as the most likely causal risk factor for stroke in the chemokine gene cluster.

Conclusions:

• The novel principal component analysis-based multivariable Mendelian randomization approach enhances causal inference from clustered genetic variants.
• Monocyte chemoattractant protein-1 is implicated as a direct causal risk factor for stroke.
• This methodology offers a robust solution for complex genetic association studies.