A unified framework for joint-tissue transcriptome-wide association and Mendelian randomization analysis

Abstract

Here we present a Joint-Tissue Imputation (JTI) approach and a Mendelian Randomization (MR) framework for causal inference, MR-JTI. JTI borrows information across transcriptomes of different tissues, leveraging shared genetic regulation, to improve prediction performance in a tissue-dependent manner. Notably, JTI includes single-tissue imputation PrediXcan as a special case and outperforms other single-tissue approaches (BSLMM and Dirichlet Process Regression). MR-JTI models variant-level heterogeneity (primarily due to horizontal pleiotropy, addressing a major challenge of TWAS interpretation) and performs causal inference with type-I error control. We make explicit the connection between the genetic architecture of gene expression and of complex traits, and the suitability of MR as a causal inference strategy for TWAS. We provide a resource of imputation models generated from GTEx and PsychENCODE panels. Analysis of biobanks and meta-analysis data and extensive simulations show substantially improved statistical power, replication, and causal mapping rate for JTI relative to existing approaches.

A systematic analysis of genetically regulated differences in gene expression and the role of co-expression networks across 16 psychiatric disorders and substance use phenotypes

“…However, it is also possible the transcriptomic correlations are inflated due to the local correlation structure of gene expression at a locus associated with two or more phenotypes. These scenarios may be investigated using recently developed computational tools for casual inference, such as FOCUS (36) or MR-JTI (37), to identify a reliable set of independent causal genes underlying each phenotype.…”

“…As such, the approach does not provide direct evidence of causal relationship between expression and disease risk. Mendelian randomisation-based approaches, such as SMR (52) and MR-JTI (37), may refine our list of gene candidates by selecting genes most likely associated through pleiotropy, where gene expression and a phenotype are affected by the same causal variant. Finally, our gene co-expression analyses rely on the stability (i.e.…”

“…Expression weights from PyschENCODE were not available at single cell resolution. Therefore, the imputed expression effects may reflect a mosaic of expression effects from multiple cell types rather than a single causal cell type, or the sharing of genetic regulation of gene expression (37). The generation of large single-cell eQTL datasets from the human brain will provide a valuable resource to disentangle cell-specific effects (50,51).…”

A large-scale transcriptome-wide association study (TWAS) of ten blood cell phenotypes reveals complexities of TWAS fine-mapping

“…A transcriptome-wide association study (TWAS) is one gene-based method which systematically investigates the association between genetically predicted gene expression and phenotypes of interest [6][7][8][9]. Here, we report results from a large TWAS of hematological measures using the PrediXcan method [6] to analyze data from 54,542 individuals of European ancestry from the Genetic Epidemiology Research on Adult Health and Aging (GERA) cohort (our discovery data set) [10] [11].…”

Disentangling genetic feature selection and aggregation in transcriptome-wide association studies

“…The importance of GReX is evidenced by the number of publications which either seek to improve its prediction accuracy or expand its applications. Researchers have refined the original ElasticNet- and Bayesian-based models of PrediXcan 6 and BSLMM 13 by integrating multiple tissues 14–16 , adding trans-eQTLs 11 , and incorporating improved Bayesian methods 8 . GReX counterparts have also been developed for LD-score 17 , polygenic risk score 18 , and fine-mapping 12 .…”