Transcriptome wide association studies: general framework and methods

Abstract

Background: Genome-wide association studies (GWAS) have succeeded in identifying tens of thousands of genetic variants associated with complex human traits during the past decade, however, they are still hampered by limited statistical power and difficulties in biological interpretation. With the recent progress in expression quantitative trait loci (eQTL) studies, transcriptome-wide association studies (TWAS) provide a framework to test for gene-trait associations by integrating information from GWAS and eQTL studies. Results: In this review, we will introduce the general framework of TWAS, the relevant resources, and the computational tools. Extensions of the original TWAS methods will also be discussed. Furthermore, we will briefly introduce methods that are closely related to TWAS, including MR-based methods and colocalization approaches. Connection and difference between these approaches will be discussed. Conclusion: Finally, we will summarize strengths, limitations, and potential directions for TWAS.

A unified quantile framework reveals nonlinear heterogeneous transcriptome-wide associations

“…Identifying which genes are responsible for these association signals, however, remains challenging. Given that most of the associated variants are found in noncoding regions and may be involved in gene regulatory mechanisms, transcriptome-wide association studies (TWAS) (Gusev et al, 2016;Gamazon et al, 2015;Zhao et al, 2021;Xie, Shan, Zhao, and Hou, Xie et al;Wainberg et al, 2019) have been proposed to leverage existing reference panels on genetic variation and gene expression data such as Genotype-Tissue Expression (GTEx) (Lonsdale et al, 2013;Consortium et al, 2020) to discover the potential target genes. Relative to GWAS, TWAS results identify significant genes, are less affected by linkage disequilibrium (Li et al, 2021), and provide more interpretable results, efficiently bringing us closer to a better understanding of the underlying causal mechanisms (Tang et al, 2021;Li et al, 2021).…”

Meta-Analysis of Transcriptome-Wide Association Studies across 13 Brain Tissues Identified Novel Clusters of Genes Associated with Nicotine Addiction

“…In short, TWAS involve training a predictive model of expression from the genotype in the reference panel, then using the trained model to predict the expression in the GWAS data, which are used to find the genes associated with the trait [ 7 , 8 ]. Various statistical methods and computational tools for implementing TWAS have been developed to date [ 9 , 10 ]. Since gene expression and eQTL regulation are tissue-dependent, TWAS are usually conducted in a tissue-specific manner.…”

Meta-Analysis of Transcriptome-Wide Association Studies Across 13 Brain Tissues Identified Novel Clusters of Genes Associated with Nicotine Addiction

“…Genotype-Tissue Expression (GTEx) [6]) which include both genotype and gene expression data as reference panel to infer association with a trait at the gene level [7]. In short, TWAS involve training a predictive model of expression from genotype in the reference panel, then using the trained model to predict the expression in the GWAS data, which are used to find the genes associated with the trait [7,8].Various statistical methods and computational tools for implementing TWAS have been developed to date [9,10]. Since gene expression and eQTL regulation are tissue-dependent, TWAS are usually conducted in tissue-specific manner.…”