Utilizing gene co-expression networks with the rat kidney TXG-MAPr tool to enhance safety assessment, biomarker identification and human translation

Toxicogenomic data represent a valuable source of biological information at molecular and cellular level to understand unanticipated organ toxicities. Weighted gene co-expression networks analysis can reduce the complexity of gene-level transcriptomic data to a set of biological response-networks useful for providing insights into mechanisms of drug-induced adverse outcomes. In this study, we have built co-regulated gene networks (modules) from the TG-GATEs rat kidney datasets consisting of time- and dose-response data for 41 compounds, including nephrotoxicants. Data from the 347 modules were incorporated into the rat kidney TXG-MAPr web tool, a user-friendly interface that enables visualization and analysis of module perturbations, quantified by a module eigengene score (EGS) for each treatment condition. Several modules annotated for cellular stress, renal injury and inflammation were statistically associated with concurrent renal pathologies, including modules that contain both well-known and novel renal biomarker genes. In addition, many rat kidney modules contain well annotated, robust gene networks that are preserved in other transcriptome datasets, suggesting that these biological networks translate to other (drug-induced) kidney injury cases. Moreover, preservation analysis of human kidney transcriptomic data provided a quantitative metric to assess the likelihood that rat kidney modules, and the associated biological interpretation, translate from non-clinical species to human. In conclusion, the rat kidney TXG-MAPr enables uploading and analysis of kidney gene expression data in the context of rat kidney co-expression networks, which could identify possible safety liabilities and/or mechanisms that can lead to adversity for chemical or drug candidates. Translational Statement Gene co-expression networks (modules) were generated using rat kidney toxicogenomics data, which reduced data complexity and retained quantitative mechanisms to enhance safety assessment. Several stress, injury and inflammation modules were statistically associated with renal pathologies, useful for biomarker identification. Moreover, many rat kidney modules contained well-annotated, robust gene-networks that were preserved in human patients transcriptome data after renal transplantation, suggesting that these biological networks translate to human relevant kidney-injury. So, the rat kidney TXG-MAPr tool enables transcriptome analysis in the context of kidney co-expression networks, which could identify chemical-induced safety liabilities and/or mechanisms leading to adversity, relevant for human risk-assessment. ### Competing Interest Statement JSR reports funding from GSK, Pfizer and Sanofi and fees/honoraria from Travere Therapeutics, Stadapharm, Astex, Owkin, Pfizer and Grunenthal. PT is a Sanofi employee and may hold shares and/or stock options in the company. All the other authors have declared no competing interests. * AKI : Acute kidney injury APL : Allopurinol ATF : Activating transcription factor AvAbsEGS : Average absolute eigengene score BUN Blood urea nitrogen corEG : Correlation eigengene score CRE : Creatinine CSP : Cisplatin DIKI : Drug-induced kidney injury DM : DrugMatrix EGS : Eigengene score ER : Endoplasmic reticulum FC : Fold change GO : Gene ontology IRI : Ischemia reperfusion injury KIM-1 : Kidney injury molecule-1 NRF2 : Nuclear Factor Erythroid 2-Related Factor 2 ORA : Over Representation Analysis PAN : Puromycin aminonucleoside PTC : Proximal tubular cells RMA : Robust Multi-array Average TG : TG-GATEs TF : Transcription factor TXG : Toxicogenomics WGCNA : Weighted gene co-expression network analysis