Transcriptome‐wide association studies accounting for colocalization using Egger regression

Abstract

Integrating genome-wide association (GWAS) and expression quantitative trait locus (eQTL) data into transcriptome-wide association studies (TWAS) based on predicted expression can boost power to detect novel disease loci or pinpoint the susceptibility gene at a known disease locus. However, it is often the case that multiple eQTL genes colocalize at disease loci, making the identification of the true susceptibility gene challenging, due to confounding through linkage disequilibrium (LD). To distinguish between true susceptibility genes (where the genetic effect on phenotype is mediated through expression) and colocalization due to LD, we examine an extension of the Mendelian randomization (MR) egger regression method that allows for LD while only requiring summary association data for both GWAS and eQTL. We derive the standard TWAS approach in the context of MR and show in simulations that the standard TWAS does not control type I error for causal gene identification when eQTLs have pleiotropic or LD-confounded effects on disease. In contrast, LD-aware MR-Egger (LDA MR-Egger) regression can control type I error in this case while attaining similar power as other methods in situations where these provide valid tests. However, when the direct effects of genetic variants on traits are correlated with the eQTL associations, all of the methods we examined including LDA MR-Egger regression can have inflated type I error. We illustrate these methods by integrating gene expression within a recent large-scale breast cancer GWAS to provide guidance on susceptibility gene identification.

Mendelian randomization accounting for correlated and uncorrelated pleiotropic effects using genome-wide summary statistics

“…If it is mean zero, it only adds noise to the relationship in Equation (1) . Multiple methods have been proposed to deal with this, including Egger regression 8 , 9 , which allows for directional (non-mean zero) uncorrelated pleiotropy, and methods that rely on outlier removal, including GSMR 10 and MR-PRESSO 7 . Correlated pleiotropy results in correlation between β M,j and β Y,j for a subset of variants.…”

Testing and controlling for horizontal pleiotropy with probabilistic Mendelian randomization in transcriptome-wide association studies

“…Some methods (e.g., Egger regression 8 , 22 , GLIDE 23 , GSMR 4 , MR-median method 9 , profile score approach 24 , MRMix 25 , and Bayesian MR 26 , 27 ) test and control for horizontal pleiotropic effects, but can only accommodate independent instruments. As far as we are aware, there is only one two-sample MR method currently developed for testing and controlling for pleiotropic effects in the presence of correlated instruments: LDA MR-Egger 28 . Unfortunately, as we will show below, LDA MR-Egger cannot handle realistic LD pattern among cis -SNPs for TWAS applications.…”

A general framework for functionally informed set-based analysis: Application to a large-scale colorectal cancer study

“…When there is only one mediator, the test statistic for testing the association of the mediator under the assumption of no direct effects has the same form as PredXcan and TWAS [5,6] and the Mendelian randomization two-stage estimator [8,13]. Our method also avoids the direct inversion of the covariance matrix of genotypes as in the Mendelian randomization approach for dealing with correlated variants [8], therefore, it can work on genes with varying degrees of correlation structures, without any variant pruning. We show that our method of combining summary statistics gives p-values that are consistent with those constructed from individual level data, and that it is much more efficient computationally.…”

“…A comprehensive review and comparison of various methods can be found in Barbeira et al (2018) [ 5 ]. The TWAS-like analysis can also be framed as a class of Mendelian randomization [ 7 , 8 ], in which under some assumptions the mediator effect of gene expression can be estimated by the inverse variance weighted ratios of regression coefficients of genetic variants for the phenotype and those for the gene expression. All of these methods also apply to other types of mediators including methylation and lifestyle variables (e.g., smoking) that may be regulated by genetic variants.…”