An integrative multi-context Mendelian randomization method for identifying risk genes across human tissues.

Area of Science:

• Genetics and Bioinformatics
• Statistical Genetics
• Causal Inference

Background:

• Mendelian randomization (MR) is crucial for assessing causal effects of exposures on outcomes.
• Conventional MR methods struggle with limited expression quantitative trait loci (eQTLs) as instrumental variables (IVs).
• Tissue-specific eQTL effects violate MR assumptions of consistent IV effects across datasets.

Purpose of the Study:

• To introduce mintMR, a novel multi-context multivariable integrative MR framework.
• To address challenges in mapping risk genes using molecular trait data.
• To improve the estimation of sparse causal effects by integrating multi-tissue eQTLs.

Main Methods:

• Developed a multi-context multivariable integrative MR framework (mintMR).
• Modeled molecular exposures across multiple tissues and gene regions simultaneously.
• Employed multi-view learning to model latent disease relevance indicators across tissues and traits.
• Iteratively performed multi-tissue MR and joint learning of tissue relevance probabilities.

Main Results:

• mintMR utilizes eQTLs with consistent effects across multiple tissues, enhancing IV consistency.
• The framework effectively models molecular trait effects across diverse biological contexts.
• Applied mintMR to 35 complex traits, evaluating gene expression and DNA methylation effects.
• Demonstrated control of genome-wide inflation and provided insights into disease mechanisms.

Conclusions:

• mintMR offers a robust framework for causal inference using multi-tissue molecular data.
• The method improves the estimation of sparse causal effects, particularly for complex traits.
• mintMR enhances our understanding of genetic and molecular contributions to disease etiology.