Testing and controlling for horizontal pleiotropy with probabilistic Mendelian randomization in transcriptome-wide...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Genome-wide association studies (GWASs) identify genetic variants associated with traits.
• Transcriptome-wide association study (TWAS) integrates gene expression data with GWASs to infer gene-trait causality.
• Existing methods face challenges with correlated instruments and horizontal pleiotropy.

Purpose of the Study:

• Introduce PMR-Egger, a probabilistic Mendelian randomization (MR) method tailored for TWAS.
• Unify existing TWAS and MR methodologies within a likelihood framework.
• Provide a scalable and robust method for causal inference of gene-trait relationships.

Main Methods:

• Developed a probabilistic Mendelian randomization (MR) likelihood framework (PMR-Egger).
• Accommodates multiple correlated genetic instruments.
• Designed to test causal gene effects in the presence of horizontal pleiotropy.

Main Results:

• PMR-Egger demonstrates calibrated type I error control in simulations with horizontal pleiotropy.
• The method is robust to model misspecifications and more powerful than existing approaches.
• PMR-Egger can directly detect horizontal pleiotropy.

Conclusions:

• PMR-Egger offers a unified, powerful, and scalable framework for TWAS.
• The method advances the investigation of molecular mechanisms underlying complex diseases.
• Demonstrated utility in analyzing 39 diseases and complex traits from UK Biobank GWAS data.