A statistical framework for cross-tissue transcriptome-wide association analysis

Abstract

Transcriptome-wide association analysis is a powerful approach to studying the genetic architecture of complex traits. A key component of this approach is to build a model to impute gene expression levels from genotypes using samples with matched genotypes and gene expression data in a given tissue. However, it is challenging to develop robust and accurate imputation models with a limited sample size for any single tissue. Here, we first introduce a multi-task learning method to jointly impute gene expression in 44 human tissues. Compared with single-tissue methods, our approach achieved an average 39% improvement in imputation accuracy and generated effective imputation models for an average 120% more genes. We then describe a summary statistic-based testing framework that combines multiple single-tissue associations into a powerful metric to quantify the overall gene-trait association. We applied our method, called UTMOST, to multiple genome wide association results and demonstrate its advantages over single-tissue strategies.

Transcriptome-wide association analysis of 211 neuroimaging traits identifies new genes for brain structures and yields insights into the gene-level pleiotropy with other complex traits

“…The present study faces some limitations. First, since these results are purely based on statistical associations, it is hard to draw conclusions about the underlying causality and prioritize causal genes 42,100 . This is also one of the main challenges for most of the current TWAS approaches 48 .…”

“…Here we applied TWAS methods to 211 structural neuroimaging traits including 101 ROI volumes and 110 DTI parameters. As these brain-related traits tend to be highly polygenic 21,36 and are related with many traits across different categories 11 , we used a cross-tissue (panel) TWAS approach (UTMOST 42 ) in our main analysis. UTMOST first performs single-tissue gene-trait association analysis in each reference panel with both within-tissue and cross-tissue statistical penalties, and then combines these single-tissue results using the Generalized Berk-Jones (GBJ) test 51 , which is aware of tissue-dependence and can account for the potential sharing of local expression regulation across tissues.…”

Influence of tissue context on gene prioritization for predicted transcriptome-wide association studies

“…It also suggested the importance of biological context in eQTL studies, and the ensemble of TWAS methods with different transcriptional regulation assumptions gave a more comprehensive picture of gene-trait relationships. In the future, we would like to perform cross-tissue TWAS analysis 12,49 , which aggregate gene-trait association information across all tissues and even across different consortia to further prioritize the trait-related genes and better describe the genetic architecture of complex diseases. Manhattan plot of gene-trait associations identified by PrediXcan.…”

“…In this study, we aimed to address whether and how single tissue and integrative tissue context of eQTLs influence TWAS gene prioritization by comparing two distinct TWAS methods, PrediXcan 11 and Unified Test for MOlecular SignaTures (UTMOST 12 ). PrediXcan uses elastic-net regression model and identifies eQTLs in a tissue-by-tissue manner.…”

Integrating germline and somatic genetics to identify genes associated with lung cancer

“…In our paper we extend the TWAS framework to include somatic information and tumor purity, thus facilitating the analysis of tumor expression data. There is evidence that the use of different tissue expression weights, and the combination of tissue expression weights, significantly affects eQTL results and TWAS association results (Gusev et al, ; Hu et al, ). Additionally, there is evidence that using tissue expression weights from tissue relevant to the trait in question may be most informative (Wainberg et al, ).…”