Differentiate Horizontal Pleiotropy from Mediation Using GWAS Summary Statistics in Combining Mendelian Randomization Analysis

摘要

The overall association evidence of a genetic variant with multiple traits can be evaluated by cross phenotype association analysis using summary statistics from individual trait GWAS. Dissecting the association pathways from a variant to multiple traits is extremely important but has not been well studied. In this study, we introduce a computationally efficient iterative approach using summary statistics from GWAS to differentiate horizontal pleiotropy from mediation in combining with Mendelian randomization analysis. Our extensive simulations suggest that the proposed iterative method has similar performance to the widely used MR-PRESSO for two-sample MR analysis but is substantially improved when using overlapped samples. Furthermore, our approach is computationally much faster than MR-PRESSO. Similar to ME-PRESSO, our proposed method leads unbiased estimates of causal effects when horizontal pleiotropy occurs in less than 50% instrumental variables. We applied our proposed method to the summary statistics from blood pressure (BP) and coronary artery disease (CAD) GWAS and detected multiple pleiotropy variants and significant causal relationships between BP and CAD.