Modern collections of transcriptional profiling experiments contain enormous wealth of information, which is severely underutilized due to inconsistent annotation, cross-platform differences and wide spectrum of conditions and tissues profiled. On the other hand, most of the modern pathway analysis tools rely on curated gene sets that quickly become outdated and often fail to capture true diversity of transcriptional responses in real biological systems. To reveal hypothesis generation potential of transcriptional profiling databases, we developed GeneQuery, new geneset-based global phenotype searching tool that makes use of gene expression data in GEO database.

GeneQuery circumvents aforementioned difficulties by introducing digital definition of phenotypes through gene modules co-expressed in a given dataset. Since there is an established connection between co-expression and co-regulation of groups of genes, we used co-expressed modules as a representative of particular phenotype in the transcriptional universe. Careful application of WGCNA approach allowed us to automatically and unbiasedly obtain co-expressed modules of genes that are subsequently compared to the geneset in question. Using regular Fisher’s exact test with Bonferroni correction, we were able to establish a phenotype search engine that finds biologically similar experiments based on the transcriptional signatures. Furthermore, using network methods we have analyzed the cross-connectivity of the overall “transcriptional universe” graphs of humans, mice, and rat, and have found that both conserved and species-specific clusters are present for each species. Overall, nearly half of all available transcriptional experiments (spanning over 400,000 samples) are included in the database, which is dynamic and easily expandable.

GeneQuery revealed an unexpected connection between transcriptional signatures of patients with Nasu-Hakola disease, a rare neurodegenerative disease caused by TREM2 and DAP12 mutations, and a portion of the aging-signature in mouse brain consisting of genes responsive to α/γ-tocopherol treatment. Utilizing a mouse model of TREM2-associated microglial deficiency, we demonstrated that α/γ-tocopherol treatment rescued microglial function in Trem2-deficient mice but did not affect WT microglia. 

GeneQuery is available free of charge at https://artyomovlab.wustl.edu/genequery/searcher/.