Predictive models and computational toxicology.

Area of Science:

• Computational toxicology
• Environmental health
• In vitro toxicology

Background:

• Assessing health risks of numerous environmental chemicals is challenging due to uncharacterized exposures and toxicities.
• The EPA's ToxCast program, part of the Tox21 consortium, aims to efficiently prioritize toxicity testing for thousands of chemicals.
• ToxCast utilizes high-throughput screening (HTS) and in vitro assays to predict chemical toxicity and inform human health assessments.

Purpose of the Study:

• To review progress in ToxCast predictive modeling for chemical toxicity.
• To evaluate the prediction of in vivo animal effects using in vitro data.
• To present lessons learned from Phase I of the ToxCast program, focusing on developmental and reproductive effects.

Main Methods:

• Utilized a workflow integrating over 650 in vitro assays (biochemical, cellular, alternative models).
• Profiled an initial library of 309 environmental chemicals, primarily pesticides with existing in vivo data.
• Developed predictive models to assess the correlation between in vitro assay results and in vivo animal study outcomes.

Main Results:

• Phase I of ToxCast successfully generated in vitro data for 309 chemicals.
• Models were built to predict in vivo animal effects based solely on in vitro screening data.
• Both ToxCast in vitro data and the ToxRefDB in vivo database are publicly accessible.

Conclusions:

• ToxCast provides a cost-effective approach for prioritizing chemical toxicity testing.
• Phase I demonstrated the feasibility of predicting in vivo effects from in vitro data.
• Ongoing Phase II efforts continue to advance predictive toxicology for environmental chemical risk assessment.