Abstract
Transcriptome-wide association studies (TWASs) are powerful for identifying gene-trait associations by integrating gene expression and genome-wide association data, but findings may be impacted by the choice of gene expression reference. We performed TWAS of cardiovascular outcomes using multi-tissue and ancestry-matched gene expression references. We used data from the Chronic Renal Insufficiency Cohort Study for participants of African (AFR, n = 1,512) and European (EUR, n = 2,067) ancestry and three outcomes: all-cause stroke, coronary heart disease, and heart failure. We performed TWASs using EUR and AFR predicted gene expression reference panels and multi-tissue TWAS by integrating gene expression from 10 GTEx selected tissues. TWAS identified KDELR2 associated with heart failure in AFR participants using matched AFR reference panel (p = 4.7 × 10(-6)), although findings were near significant using the EUR mismatched reference panel (p = 5.6 × 10(-6)). PSMC1 was associated with coronary heart disease in TWAS of CRIC EUR using AFR reference panel, and this gene was not present in the EUR-trained gene expression model. Multi-tissue TWASs identified KHDRBS2 significantly associated with all-cause stroke in CRIC AFR participants (p = 4.0 × 10(-6)). Variants near KDELR2 have been associated with coronary artery disease, which is a main cause of heart failure, while KHDRBS2 has been associated with cardiovascular risk factors in genome-wide association studies. Our findings highlight differences in gene discovery for TWAS of cardiovascular disease applied to high-risk participants based on participant ancestry and gene expression reference panels, and gains to identify genes compared with traditional genome-wide association approaches.