MRAID

MRAID performs two-sample Mendelian Randomization using GWAS summary statistics to estimate causal effects between traits while accounting for horizontal pleiotropy and automatically selecting instrumental SNPs using fine-mapping concepts.

Key Features:

• Two-sample MR with GWAS summary statistics: Uses genome-wide association study summary statistics to estimate causal effects of one trait on another.
• Automated instrument determination via fine mapping: Begins with candidate SNPs potentially in high linkage disequilibrium and automatically selects suitable instruments informed by fine-mapping concepts.
• Modeling of correlated and uncorrelated horizontal pleiotropy: Explicitly models both uncorrelated and correlated horizontal pleiotropic effects to reduce bias from pleiotropy.
• Joint likelihood framework: Performs inference within a joint likelihood framework to integrate instrument selection and effect estimation.
• Scalable sampling-based algorithm for calibrated p-values: Utilizes a scalable sampling-based algorithm to compute calibrated p-values for hypothesis testing.
• Calibrated type I error control and improved power: Achieves calibrated type I error control and enhanced power relative to some existing methods as demonstrated in simulations.

Scientific Applications:

• Causal inference between traits and diseases: Estimates causal relationships between traits and disease outcomes using summary-level genetic data.
• Large-scale MR screening in UK Biobank: Applied in an MR screening of 645 trait pairs from the UK Biobank to identify lifestyle-related causal risk factors for cardiovascular disease traits.

Methodology:

Integrates fine-mapping concepts to select instruments from candidate SNPs in high linkage disequilibrium, explicitly models correlated and uncorrelated horizontal pleiotropy, employs a joint likelihood framework, and uses a scalable sampling-based algorithm to compute calibrated p-values.