Toxicogenomic module associations with pathogenesis: a network-based approach to understanding drug toxicity

Abstract

Despite investment in toxicogenomics, nonclinical safety studies are still used to predict clinical liabilities for new drug candidates. Network-based approaches for genomic analysis help overcome challenges with whole-genome transcriptional profiling using limited numbers of treatments for phenotypes of interest. Herein, we apply co-expression network analysis to safety assessment using rat liver gene expression data to define 415 modules, exhibiting unique transcriptional control, organized in a visual representation of the transcriptome (the 'TXG-MAP'). Accounting for the overall transcriptional activity resulting from treatment, we explain mechanisms of toxicity and predict distinct toxicity phenotypes using module associations. We demonstrate that early network responses complement traditional histology-based assessment in predicting outcomes for longer studies and identify a novel mechanism of hepatotoxicity involving endoplasmic reticulum stress and Nrf2 activation. Module-based molecular subtypes of cholestatic injury derived using rat translate to human. Moreover, compared to gene-level analysis alone, combining module and gene-level analysis performed in sequence identifies significantly more phenotype-gene associations, including established and novel biomarkers of liver injury.

Crosstalk between Receptor and Non-receptor Mediated Chemical Modes of Action in Rat Livers Converges through a Dysregulated Gene Expression Network at Tumor Suppressor Tp53

“…Notably scoring of the modules using the Nextbio Body Atlas (Data not shown) reveals that genes in all four modules are over expressed in inflammatory cells including the blood, suggesting that what may be being detected are transcripts from inflammatory infiltrates that manifest following tissue damage. Notably this is consistent with observations that co-regulation modules in the liver are related, in part to changes in cellularity (i.e., increases or decreases in certain cell types; Sutherland et al, 2017 ). Chemicals that act through a receptor have cascades of signaling that often attenuate cell death by inhibiting apoptosis (Mally and Chipman, 2002 ) therefore decreasing cell turnover.…”

“…The approach uses transformations of the correlation between co-expressed genes to reveal interconnectedness amongst gene network nodes and permutation procedures to identify statistically significant gene network modules that overlap between sample conditions (Langfelder et al, 2011 ). Recent utilization of WGCNA on rat liver gene expression data from drug toxicity studies revealed 415 gene network models that associate with mechanisms of liver pathogenesis (Sutherland et al, 2017 ). We used the WGCNA approach to reconstruct gene networks using microarray gene expression data from male rat livers and identify preserved modules between chemicals that exert their MOA through receptors (RM) vs. those that are non-receptor-mediated (NRM).…”

Transcriptomics in Toxicogenomics, Part III: Data Modelling for Risk Assessment

“…Gene co-expression network analysis is a systems biology method used to describe the correlation patterns among genes across different experimental samples. It allows representing, investigating and understanding the complex molecular interactions within the exposed system [22,45]. The genes and their interactions are represented as a network (or graph) where the genes are the nodes of the network and their strength of similarity is represented as weighted edges between the nodes [46].…”

“…Subsequently, they ranked the genes in the network by computing different topological measures, identified and functionally characterized a set of genes that play a key role in the adaptation to exposure. Other approaches focus on identifying gene network modules associated with specific patterns of drug toxicity [45].…”

“…These methods differ on how they calculate the similarities between the expression profiles and how they remove the non-relevant connections. The dependence between the expression profiles is usually computed by means of information-theoretic methods such as the pairwise correlation coefficient and mutual information (MI) [45,55]. The main difference between the two methods is that the first one is able to identify only linear dependence between the profiles, while the second is also able to identify non-linear dependencies.…”