TWAS-CTL: A robust and efficient method for multi-tissue transcriptome-wide association studies using cross-tissue learners

The advent of transcriptome-wide association studies (TWAS) has expanded the classical genome-wide association study (GWAS) framework by integrating gene expression with genetic variation to identify trait-associated variants. While multi-tissue TWAS approaches improve statistical power over single-tissue models, existing methods often lose information during result aggregation and require intensive computation. Here, we present TWAS-CTL (Cross Tissue Learner), a novel framework that leverages heterogeneous gene expression across tissues by adaptively reweighting and optimizing multiple single-tissue learners. Simulations demonstrate that TWAS-CTL achieves higher statistical power than the leading method, UTMOST, while maintaining proper type I error control and reducing computational time by over half. When applied to the analysis of the Genetics of Kidneys in Diabetes (GOKIND) cohort, we observed that TWAS-CTL identified more susceptible genes associated with diabetes than both UTMOST and PrediXcan, another widely used method in TWAS. These results establish TWAS-CTL as a powerful and efficient tool for cross-tissue gene expression analysis, capable of integrating heterogeneous gene-trait associations to advance genetic discovery.