Aggregating multiple expression prediction models improves the power of transcriptome-wide association studies.

Area of Science:

• Genetics
• Bioinformatics
• Statistical genomics

Background:

• Transcriptome-wide association study (TWAS) integrates genetic and gene expression data to identify genes linked to phenotypes.
• Existing TWAS methods use various prediction models, each suited for specific genetic architectures of gene expression.
• The diversity of gene expression genetic architectures necessitates flexible TWAS approaches.

Purpose of the Study:

• To develop a novel TWAS method, HMAT, that aggregates association evidence across multiple gene expression prediction models.
• To improve the power and accuracy of TWAS by accounting for varied genetic architectures.
• To identify novel disease-associated genes by enhancing TWAS analysis.

Main Methods:

• HMAT employs a harmonic mean P-value combination strategy to aggregate TWAS results from diverse prediction models.
• The method accounts for correlations among TWAS test statistics to ensure calibrated P-values.
• Performance was evaluated through numerical simulations and application to summary statistics of nine common diseases.

Main Results:

• HMAT demonstrated superior performance compared to existing TWAS methods and ad hoc P-value combination strategies in simulations.
• In real data analyses, HMAT achieved an average 30.6% increase in power over the next best method.
• HMAT identified numerous novel disease-associated genes missed by conventional TWAS approaches.

Conclusions:

• HMAT is a flexible and powerful TWAS method effective across various genetic architectures of gene expression.
• The harmonic mean-based aggregation strategy enhances gene discovery for complex traits.
• HMAT offers a robust approach for advancing genetic association studies.