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

Area of Science:

• Toxicogenomics
• Systems Biology
• Computational Biology

Background:

• Nonclinical safety studies are crucial for predicting drug clinical liabilities.
• Whole-genome transcriptional profiling faces challenges with limited treatment data.
• Network-based genomic analysis offers a solution for complex safety assessments.

Purpose of the Study:

• To apply co-expression network analysis to rat liver gene expression data for safety assessment.
• To develop a visual representation of the transcriptome (TXG-MAP) organizing gene modules.
• To predict toxicity mechanisms and phenotypes using network analysis.

Main Methods:

• Co-expression network analysis of rat liver gene expression data.
• Definition and organization of 415 gene modules into the TXG-MAP.
• Integration of module associations with transcriptional activity to explain toxicity.
• Comparison of network-based analysis with traditional histology and gene-level analysis.

Main Results:

• Identified 415 gene modules with unique transcriptional control within the TXG-MAP.
• Demonstrated that early network responses predict outcomes in longer studies.
• Discovered a novel hepatotoxicity mechanism involving endoplasmic reticulum stress and Nrf2 activation.
• Showcased translation of rat-derived cholestatic injury subtypes to human data.
• Found that combined module and gene-level analysis significantly enhances phenotype-gene association identification.

Conclusions:

• Network-based genomic analysis, specifically co-expression networks and the TXG-MAP, improves nonclinical safety assessment.
• Early network responses provide valuable predictive information complementing traditional methods.
• The approach identifies novel toxicity mechanisms and biomarkers, with translatable findings between species.
• Combining module and gene-level analysis is superior to gene-level analysis alone for identifying phenotype-gene associations.