Bayesian Mendelian Randomization Analysis for Latent Exposures Leveraging GWAS Summary Statistics for Traits...

Area of Science:

• Genetics
• Epidemiology
• Statistical genetics

Background:

• Mendelian Randomization (MR) infers causality from observational data using genetic variants as instrumental variables.
• A challenge arises when exposures are unobserved latent factors regulating correlated traits.
• Existing MR methods struggle with jointly analyzing multiple correlated traits influenced by unobserved exposures.

Purpose of the Study:

• To develop a Bayesian MR framework for jointly analyzing latent exposures influencing multiple correlated traits.
• To assess causal effects of unobserved biological factors on disease outcomes using GWAS summary statistics.
• To improve causal inference in complex genetic and epidemiological studies.

Main Methods:

• Proposed a Bayesian MR framework for joint analysis of multiple latent exposures.
• Leveraged genome-wide association study (GWAS) summary-level statistics for co-regulated traits.
• Employed simulation studies to evaluate method performance against alternatives.

Main Results:

• The proposed Bayesian MR framework demonstrated superior type I error control and statistical power.
• The method showed improved validity and stability compared to traditional and alternative approaches.
• Identified potential causal links between psychiatric factors and various diseases.

Conclusions:

• The novel Bayesian MR framework effectively analyzes causal effects of latent exposures on disease outcomes.
• This method offers a robust approach for complex trait analysis in genetic epidemiology.
• Provides evidence for causal roles of psychiatric factors in diverse disease categories.