Integrating Concentration-Dependent Toxicity Data and Toxicokinetics To Inform Hepatotoxicity Response Pathways.

Area of Science:

• Computational toxicology
• Chemical safety assessment
• In vitro toxicology

Background:

• Animal models often fail to predict human liver toxicity.
• In vitro high-throughput screening (HTS) assays offer an alternative but require pathway-based models for complex toxicities.
• Existing models for simple pathways are successful, but complex toxicities like hepatotoxicity remain challenging.

Purpose of the Study:

• To develop a computational strategy for creating pathway-based models for complex toxicities, specifically human hepatotoxicity.
• To identify and group in vitro assays relevant to hepatotoxicity mechanisms.
• To improve the prediction of in vivo hepatotoxicity using these models.

Main Methods:

• Utilized a database of 2171 chemicals with human hepatotoxicity classifications.
• Screened 1600+ ToxCast/Tox21 HTS assays to identify those associated with hepatotoxicity.
• Developed a computational framework to group assays into 52 key event (KE) models based on biological targets/mechanisms.
• Employed supervised learning with KE scores and toxicokinetic information for prediction.

Main Results:

• Identified 157 HTS assays linked to human hepatotoxicity.
• Grouped assays into 52 KE models, generating KE scores for chemical potency.
• Observed that chemical structure groupings with high KE scores plausibly indicated hepatotoxicity mechanisms.
• Achieved improved prediction of in vivo hepatotoxicity by integrating KE scores and toxicokinetic data.

Conclusions:

• The developed computational strategy provides a universal approach for pathway-based modeling of complex toxicities.
• This method enhances the prediction of chemical-induced liver injury.
• The strategy holds potential for broader applications in chemical toxicity evaluations.