CCmed: cross-condition mediation analysis for identifying robust trans-eQTLs and assessing their effects on human traits

Abstract

Trans-eQTLs collectively explain a substantial proportion of expression variation, yet are challenging to detect and replicate since their effects are individually weak. Many trans-effects are mediated by cis-gene expression and some of those effects are shared across tissue types/conditions. To detect robust cis-mediated trans-associations at the gene-level and for specific single nucleotide polymorphisms (SNPs), we proposed two Cross-Condition Mediation methods -CCmed gene and CCmed GWAS , respectively. We analyzed data from 13 brain tissue types from the Genotype-Tissue Expression (GTEx) project, and identified trios with cis-eQTLs of a cis-gene having associations with a trans-gene, many of which show evidence of replication in other datasets. By applying CCmed GWAS , we identified trans-genes associated with known schizophrenia susceptibility loci. We further conducted validation analyses assessing the schizophrenia-risk-associations of the identified trans-genes, by harnessing GWAS summary statistics from the Psychiatric Genomics Consortium and multitissue eQTL statistics from GTEx. Availability of data and materialThe R package CCmed is available at https://github.com/kjgleason/CCmed. The R package MrRobin is available at https://github.com/kjgleason/MrRobin.Author's contributions LSC and FY conceived the project and developed the methods. LSC, FY and KJG wrote the manuscript. LSC and KJG analyzed the data. JW, JD, XH and BLP provided valuable suggestions to the development of the methods and analysis. KJG and FY developed the R software package.

Co-expression analysis reveals interpretable gene modules controlled by trans-acting genetic variants

“…This suggests that if cis and trans effects are separated from each other either in time (early versus late response) or space (different cell types that interact with each other), then this might limit the power of methods that rely on genetically predicted gene expression levels to identify regulatory interactions ( Liu et al, 2018 ; Luijk et al, 2018 ; Wheeler et al, 2019 ) and infer causal models. This can also have a negative impact on mediation analysis ( Battle et al, 2014 ; Chick et al, 2016 ; Yang et al, 2019 ), which seeks to estimate the proportion of trans -eQTL variance explained by the expression level of the cis gene. Altogether, our results indicate that limiting trans -eQTL analysis to missense variants and to variants that have been detected as cis -eQTLs in the same cell type might miss some true associations, because the cis effect might be active in some other, yet unprofiled, context.…”

Co-expression analysis reveals interpretable gene modules controlled bytrans-acting genetic variants

“…This suggests that if cis and trans effects are separated from each other either in time (early versus late response) or space (different cell types that interact with each other), then this might limit the power of methods that rely on genetically predicted gene expression levels to identify regulatory interactions [22,48,49] and infer causal models. This can also have a negative impact on mediation analysis [50][51][52], which seeks to estimate the proportion of trans-eQTL variance explained by the expression level of the cis gene. Altogether, these results indicate that limiting trans-eQTL analysis to missense variants and to variants that have been detected as cis-eQTLs in the same cell type might miss some true associations, because the cis effect might be active in some other, yet unprofiled, context.…”

DeCompress: tissue compartment deconvolution of targeted mRNA expression panels using compressed sensing

“…This interaction term's effect size may also be inflated for compartment estimates that have low mean and high variance across the samples. In addition, we did not consider trans-acting eQTLs that are often attributed to compartment heterogeneity, though we believe that methods employing mediation or cross-condition analysis can be integrated with compartment estimates to map compartment-specific trans-eQTLs relevant in breast cancer (114)(115)(116).…”