Reciprocal causation mixture model for robust Mendelian randomization analysis using genome-scale summary data.

Area of Science:

• Genetics
• Epidemiology
• Statistical genetics

Background:

• Mendelian randomization (MR) is widely used for causal inference from GWAS summary statistics.
• Pleiotropy in GWAS complicates MR by potentially violating its core assumptions.
• Existing MR methods typically analyze causation in only one direction, requiring separate analyses for bidirectional inference.

Purpose of the Study:

• To develop a novel statistical framework, MRCI (Mixture model Reciprocal Causation Inference), for simultaneous bidirectional causal inference.
• To address challenges posed by pleiotropy in genetic association studies.
• To enable robust causal inference between two phenotypes using genome-scale summary statistics.

Main Methods:

• MRCI utilizes genome-scale summary statistics from two phenotypes and reference linkage disequilibrium (LD) information.
• The framework employs a mixture model approach to estimate reciprocal causation.
• It is designed to handle correlated pleiotropy and perform simultaneous bidirectional analyses.

Main Results:

• Simulation studies demonstrated that MRCI provides nearly unbiased causal estimates in both directions, even with strong correlated pleiotropy.
• MRCI maintained correct Type I error rates under the null hypothesis in simulations.
• Application to real GWAS data identified significant bidirectional and unidirectional causal influences between common diseases and risk factors.

Conclusions:

• MRCI offers a robust statistical framework for estimating reciprocal causation between phenotypes.
• The method effectively handles pleiotropy and allows for simultaneous bidirectional causal inference.
• MRCI enhances the ability to infer complex causal relationships from large-scale genetic data.