Causal effect heterogeneity estimation using summary statistics.

Area of Science:

• Genetic Epidemiology
• Statistical Genetics
• Causal Inference

Background:

• Mendelian randomization (MR) is a popular tool for causal inference in genetic epidemiology.
• Current MR methods primarily estimate average causal effects, lacking the ability to quantify heterogeneity.
• This limitation hinders context-dependent causal findings and a deeper understanding of complex diseases.

Purpose of the Study:

• Introduce MEndelian Randomization for Linear INteraction (MERLIN), a novel Bayesian framework.
• Jointly estimate average and context-dependent causal effects using summary-level data.
• Address the methodological limitation of existing MR approaches in quantifying causal heterogeneity.

Main Methods:

• Developed MERLIN, a unified Bayesian framework for causal inference.
• Utilized summary data from genome-wide association studies (GWAS) and interaction studies.
• Performed extensive simulation analyses to assess MERLIN's performance.

Main Results:

• MERLIN demonstrated improved power, robustness, and utility compared to existing methods.
• Identified sex-specific causal effects of schizophrenia on brain imaging traits.
• Detected a male-specific causal effect of testosterone on bipolar disorder and age-dependent effects of metabolic biomarkers on coronary artery disease risk.

Conclusions:

• MERLIN provides a powerful and practical framework for investigating causal effect heterogeneity.
• Enables summary-data-based inference for context-dependent causal relationships.
• Significantly enhances the capability to elucidate complex disease etiology and genetic epidemiology.