Analysis of Inter-tissue co-expression network using linear mixed models

Background: Inter-tissue molecular interactions are critical to the function and behavior of biological systems in multicellular organisms, but systematic studies of interactions between tissues are lacking. Also, existing studies of inter-tissue interactions are based on direct gene expression correlations, which can’t distinguish correlations due to common genetic architectures versus biochemical or molecular signal exchange between tissues.

Methods: We developed a novel strategy to study inter-tissue interaction by removing effects of genetic regulation of gene expression (genetic decorrelation) using linear mixed modesl. We applied our method to the comprehensive atlas of gene expression across nine human tissues in the Genotype-Tissue Expression (GTEx) project to generate novel genetically decorrelated inter-tissue networks. From this we derived modules of genes important in inter-tissue interactions that are likely driven by biological signal exchange instead of their common genetic basis. Importantly we highlighted communication between tissues and elucidated gene activities in one tissue inducing gene expression changes in others.

Results: We reveal global unidirectional inter-tissue coordination of specific biological pathways such as protein synthesis. Using our data, we highlighted a clinically relevant example whereby heart expression of DPP4 was coordinated with a gene expression signature characteristic for whole blood proliferation, potentially impacting peripheral stem cell mobilization.

Conclusions: In summary, this is the first resource of multi-tissue networks enabling the investigation of molecular inter-tissue interactions. With the networks in hand, we may systematically design combination therapies that simultaneously target multiple tissues or pinpoint potential side effects of a drug in other tissues.