Clinical and Therapeutic Applications of Individual-level Tissue-Specific Imputed Transcriptomes

Abstract

Gene expression causally contributes to phenotypic variability. However, since tissue-specific gene expression is difficult to measure at scale, inferring phenotypic risk for individuals using expression profiles can be difficult. We statistically imputed tissue-specific gene expression for 486,452 UK Biobank participants and demonstrated its applications using rich phenotypic data in the biorepository. We performed joint analysis of transcriptome-wide association studies (TWAS) and colocalization for six traits to prioritize tissue-gene pairs for therapeutics development. Using genes identified in TWAS, we constructed imputed transcriptome risk scores and showed their predictive utility even when conditioning on genome-wide polygenic risk scores. A phenome-wide association analysis for established therapeutic targets in whole blood recapitulated known biology and prioritized repurposing potentials supported by previous literature. Our study provides evidence supporting the utility of individual-level imputed transcriptomes for discovering novel gene-tissue-phenotype trios, expression-based phenotypic risk prediction, and identifying individuals with therapeutically targetable predisposition to disease.