S2: Modeling the Tox21 10K chemical activity profiles

Modeling the Tox21 10K chemical activity profiles for in vivo toxicity prediction

Modeling the Tox21 10K chemical activity profiles, OpenTox USA 2018

Ruili Huang




Lead Informatics Scientist


Target-specific, mechanism-oriented in vitro assays post a promising alternative to traditional animal toxicology studies. The Tox21 program, a large-scale in vitro chemical toxicity screening effort, has tested approximately 10,000 chemicals in triplicates at 15 concentrations against a panel of nuclear receptor and stress response pathway assays, producing more than 85 million data points to date that form a rich set of compound in vitro activity profiles. Compound activity data were applied in combination with structural information to build predictive models for a number of in vivo toxicity endpoints including adverse drug effects (ADE) observed in humans. Models built with human in vitro assay data achieved performance close to models based on animal in vivo toxicity data in predicting human ADEs. Combining structure and activity data resulted in better models than using structure or activity data alone. Furthermore, expanding the biological space coverage of assays by including additional drug-target annotations was shown to significantly improve model performance. A small set of targets, when added to the current suite of Tox21 assays, resulted in models that greatly outperformed those built with the existing animal toxicity data.