Transcriptome-wide association studies (TWAS) integrate expression quantitative trait loci 1 (eQTLs) studies with genome-wide association studies (GWASs) to prioritize candidate target 2 genes for complex traits. Several statistical methods have been recently proposed to improve 3 the performance of TWAS in gene prioritization by integrating transcriptome information 4 imputed from multiple tissues, and made significant achievements in improving the ability 5 to detect gene-trait associations. The major limitation of these methods is that they cannot 6 be used to elucidate the specific functional effects of candidate genes across different tissues. 7 Here, we propose a tissue-specific collaborative mixed model (TisCoMM) for TWAS, using a 8 * Correspondence should be addressed to Jin Liu (jin.liu@duke-nus.edu.sg) 1 computational framework that integrates multi-tissue eQTL analysis. TisCoMM leverages the 9 co-regulation of genetic variations across different tissues explicitly via a unified collaborative 10 mixed model. With the probabilistic framework, TisCoMM not only performs hypothesis testing 11 to prioritize gene-trait associations but also ascertains the tissue-specific role of candidate 12 target genes in complex traits. To make use of widely available GWAS summary statistics, we 13 extend TisCoMM to use summary-level data, namely, TisCoMM-S 2 . Using extensive simulation 14 studies, we demonstrate both the type I error control and the improved statistical power for 15 the proposed methods. We further illustrate the benefits of our methods in applications to 16 summary-level GWAS data for 35 traits. Notably, apart from better identifying potential 17 trait-associated genes, we can elucidate the tissue-specific role of candidate target genes. The 18 follow-up pathway analysis from tissue-specific genes for asthma shows that the immune system 19 plays an essential function for asthma development in both thyroid and lung tissues.
20Over the last decade, GWASs have achieved remarkable successes in identifying genetic 22 susceptible variants for a variety of complex traits [1]. However, the biological mechanisms to 23 understand these discoveries remain largely elusive as majority of these discoveries are located in 24 non-coding regions [2]. Recent expression quantitative trait loci (eQTLs) studies indicate that 25 the expression regulatory information may play a pivotal role bridging both genetic variants 26 and traits [3, 4, 5]. Cellular traits in comprehensive eQTL studies can serve as reference data, 27 providing investigators with an opportunity to examine the regulatory role of genetic variants on 28 gene expression. For example, the Genotype-Tissue Expression (GTEx) Project [6] has provided 29 DNA sequencing data from 948 individuals and collected gene-expression measurements of 54 30 tissues from these individuals in the recent V8 release.
31Transcriptome-wide association studies (TWAS) has been widely used to integrate the 32 expression regulatory information from these eQTL studies with GWAS to ...