MR-Corr2

MR-Corr2 accounts for correlated horizontal pleiotropy to improve causal effect estimation in two-sample Mendelian randomization using GWAS summary statistics.

Key Features:

• Two-Sample MR Methodology: Tailored for two-sample Mendelian randomization using GWAS summary statistics to analyze exposure–outcome associations without individual-level data.
• Handling Correlated Horizontal Pleiotropy: Accounts for correlated horizontal pleiotropy by reparameterizing the bivariate normal distribution of genetic effects on exposure (γ) and pleiotropy (α) via orthogonal projection within a Bayesian framework.
• Bayesian Approach with Spike-Slab Prior: Uses a spike-slab prior to distinguish pleiotropic variants from non-pleiotropic variants and reduce bias in causal estimates.
• Efficient Computational Algorithm: Implements paralleled Gibbs sampling to perform posterior inference and scale to large GWAS datasets.
• Simulation Studies for Validation: Validated by simulation studies demonstrating improved type-I error control and more accurate point estimates compared with existing methods across scenarios.

Scientific Applications:

• Causal Inference in Complex Traits: Estimate causal effects between exposures and outcomes in complex traits, exemplified by analyses of HDL cholesterol (HDL-c) and coronary artery disease (CAD).
• Understanding Causal Networks: Disentangle causal networks among complex traits to aid interpretation of biological pathways and identification of potential intervention targets.

Methodology:

Uses two-sample MR with GWAS summary statistics, reparameterizes the bivariate normal of γ and α via orthogonal projection, applies a Bayesian framework with a spike-slab prior, and performs posterior inference via paralleled Gibbs sampling; validation performed by simulation studies.