Cross-condition mediation analysis for mapping robust trans-associations and validation via two-sample Mendelian Randomization analysis

Trans-eQTLs explain a substantial proportion of expression variation, yet are challenging to be detected and replicated since their effects often act in a tissue-specific manner. Many trans-effects are mediated via cis-gene expression and some of those effects are shared across tissue types/conditions. In order to detect robust cis-mediated trans-associations, we proposed a Cross-Condition Mediation method (CCmed-gene). We analyzed data from multiple brain tissue types of the Genotype-Tissue Expression (GTEx) project, and identified 9,631 cis- and trans-gene pairs with gene-level trans-association and mediation effects, many of which showed evidence of replication in other datasets. In order to detect trans-genes for GWAS SNPs of a complex trait, we further developed CCmed-GWAS, and applied it to identify suspected trans-genes associated with 108 known schizophrenia susceptibility loci. To validate the trait-associations of the suspected trans-genes, we conducted validation analyses using a newly proposed two-sample Mendelian Randomization method, MR-Robin, in which we harnessed GWAS summary statistics from the Psychiatric Genomics Consortium and multitissue eQTL statistics from GTEx.