OTTERS: A powerful TWAS framework leveraging summary-level reference data [article]

Qile Dai, Geyu Zhou, Hongyu Zhao, Urmo Vosa, Lude Franke, Alexis Battle, Alexander Teumer, Terho Lehtimaki, Olli Raitakari, Tonu Esko, Michael P. Epstein, Jingjing Yang (+1 others)
2022 bioRxiv   pre-print
Transcriptome-wide association studies (TWAS) identify genes that influence complex traits through genetic regulation of gene expression. Existing TWAS tools like FUSION require individual-level eQTL reference data from a source like GTEx (n=~100s) to impute gene expression in a test GWAS. Therefore, these TWAS tools are not applicable to enormous summary-level reference eQTL datasets like those generated by the eQTLGen consortium (n=~32K). Development of TWAS methods that can harness
more » ... vel reference data like eQTLGen is valuable not only to enable TWAS in more general settings but also to permit more powerful analyses, since we expect expression prediction to improve with increasing reference sample size. To fill this important gap, we develop a TWAS framework called OTTERS (Omnibus Transcriptome Test using Expression Reference Summary data) that adapts multiple polygenic risk score (PRS) methods to estimate eQTL weights from summary-level eQTL reference data. For each PRS method, OTTERS estimates eQTL weights to impute gene expression per gene and then tests for association between imputed expression and outcome for a GWAS dataset. OTTERS then combines the PRS-based association tests together to create an optimal TWAS p-value. By simulation, we show that OTTERS is powerful across a wide range of models. We also applied OTTERS to conduct a TWAS of cardiovascular disease using GWAS summary data from UK Biobank and summary-level eQTLGen reference data. OTTERS identified 11 additional TWAS risk genes that were missed by FUSION using GTEx reference data. These findings suggest that OTTERS provides a practical and powerful tool for TWAS analysis.
doi:10.1101/2022.03.30.486451 fatcat:qis3sedlvrhpxdhstul2pbnrja