A tissue-specific collaborative mixed model for jointly analyzing multiple tissues in transcriptome-wide association...

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Transcriptome-wide association studies (TWASs) integrate expression quantitative trait loci (eQTLs) and genome-wide association studies (GWASs) for gene prioritization.
• Existing multi-tissue TWAS methods prioritize genes but struggle to infer tissue-specific functional effects.
• There is a need for methods that can leverage multi-tissue expression data for both gene prioritization and tissue-specific inference.

Purpose of the Study:

• To propose a novel tissue-specific collaborative mixed model (TisCoMM) for TWASs.
• To develop a summary-level data version (TisCoMM-S2) for broader application with existing GWAS data.
• To enable the detection of tissue-specific roles of candidate genes in complex traits.

Main Methods:

• Developed TisCoMM, a unified probabilistic model leveraging co-regulation of genetic variations across tissues.
• Extended TisCoMM to TisCoMM-S2 for utilizing summary-level GWAS data.
• Conducted extensive simulation studies to evaluate performance.

Main Results:

• TisCoMM-S2 controls type I error at nominal levels and significantly improves statistical power for identifying associated genes.
• The method effectively controls the false-positive rate for non-causal tissues.
• Applications to 33 complex traits identified potential trait-associated genes and elucidated their tissue-specific roles, exemplified by asthma and the immune system's function in thyroid and lung tissues.

Conclusions:

• TisCoMM and TisCoMM-S2 offer a powerful framework for TWASs, enhancing gene prioritization and enabling tissue-specific functional inference.
• These methods advance the understanding of genetic architecture of complex traits by revealing tissue-specific gene effects.
• The findings highlight the importance of considering tissue context in genetic association studies.