Leveraging expression from multiple tissues using sparse canonical correlation analysis and aggregate tests improves...

Area of Science:

• Genetics and Genomics
• Statistical Genetics
• Bioinformatics

Background:

• Transcriptome-wide association studies (TWAS) link traits to genetically predicted gene expression.
• TWAS power is limited by small sample sizes of expression quantitative trait locus (eQTL) data and lack of relevant tissue data.
• Current methods struggle to effectively integrate multi-tissue eQTL information for TWAS.

Purpose of the Study:

• To propose and evaluate a novel method for integrating multiple tissues in TWAS.
• To enhance the power and sensitivity of TWAS for detecting gene-trait associations.
• To maintain robust control over Type I error rates.

Main Methods:

• Developed a sparse canonical correlation analysis (sCCA) framework to integrate multi-tissue eQTL data.
• Combined sCCA-derived cross-tissue genetic predictors with an aggregate Cauchy association test (ACAT) for TWAS.
• Evaluated performance through simulations and application to summary statistics from 10 complex traits.

Main Results:

• The sCCA-TWAS with ACAT (sCCA+ACAT) significantly outperformed traditional single-tissue TWAS in simulation studies.
• sCCA+ACAT demonstrated higher power to detect gene-phenotype associations, especially when causal tissue expression was not directly measured.
• Application to real data identified an average of 400 additional gene-trait associations per trait compared to single-trait TWAS.

Conclusions:

• Aggregating eQTL data across multiple tissues using sCCA substantially improves TWAS sensitivity.
• The sCCA+ACAT method offers a powerful and reliable approach for gene-trait association discovery.
• This approach increases the number of testable genes and identifies novel associations missed by conventional methods.