Transcriptional Risk Scores from GWAS and EQTLs for Human Disease Prediction

• Activity Number: 176 - Statistical Genetics III – Predictive Modeling, GxE Interaction, and Causal Inference
• Type: Contributed
• Date/Time: Tuesday, August 4, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Section on Statistics in Genomics and Genetics
• Author(s): Nayang Shan* and Yuhan Xie and Shuang Song and Zuoheng Wang and Lin Hou
• Companies: Tsinghua University and Yale University and Tsinghua University and Yale University and Tsinghua University
• Keywords: gene imputation; transcriptional risk scores; risk prediction

Abstract

Polygenetic risk score (PRS) has been used for disease risk prediction. Many studies incorporated external information such as linkage disequilibrium, functional annotation, and pleiotropy among multiple diseases, to adjust the linear weights of PRS. Other than the genomic features, increasing amount of multi-omic data has motivated us to integrate multi-dimensional information. We developed a novel flexible transcriptional risk score (TRS), in which mRNA expression levels were imputed and weighted for risk prediction. We assessed the performance of the proposed method in a single tissue and multiple tissues. Applied to seven traits in the Wellcome Trust Case Control Consortium datasets as well as meta-GWAS summaries for type 2 diabetes (T2D) and Crohn’s disease (CD), we found that our method achieved better prediction accuracy than LDpred, especially for type 1 diabetes (T1D). Moreover, our method can be easily extended to epigenomic and proteomic data when reference data becomes available.