You are here:
Predicting In Vivo Effect Levels for Repeat Dose Systemic Toxicity using Chemical, Biological, Kinetic and Study Covariates
Truong, L., G. Ouedraogo, L. Pham, J. Clouzeau, S. Loisel-Joubert, D. Blanchet, H. Noçairi, Woodrow Setzer, R. Judson, Chris Grulke, K. Mansouri, AND M. Martin. Predicting In Vivo Effect Levels for Repeat Dose Systemic Toxicity using Chemical, Biological, Kinetic and Study Covariates. Archives of Toxicology. Springer, New York, NY, 92(2):587-600, (2018).
• Agency Research Drivers - EPA Program and Regional Offices are often tasked with addressing the potential hazard(s) of chemicals for which little-to-no data exist. To address this knowledge gap, the Agency initiated the ToxCast research program to generate data on a large number of chemicals using high-throughput screening methods and computational toxicology. • Science Challenge – Data from single high-throughput assays can be difficult to interpret in risk decisions. Using these data for Agency risk decisions requires contextualizing data from in vitro assays with adverse outcomes in vivo. • Research Approach – This work will use statistical and biological models of in vitro data to generate predictive signatures consistent with the likelihood of a chemical to cause systemic toxicity in vivo. These models will also generate probability estimates for the predictions of in vivo toxicity. • Results – Systemic effects were modeled using study variables (e.g., study type, strain, administration route), chemical, in vitro bioactivity, and kinetic datasets. A consensus model was developed using a combination of study-level covariates, chemical, biological, and kinetic descriptors. A benchmark model (i.e., upper expectation of model performance) was also developed by incorporating study-level covariates and the mean effect level per chemical. Although biological descriptors did not improve model performance, the final model was enriched for biological descriptors to account for kinetics and non-specific bioactivity in predicting systemic effect levels. These methods generate an externally predictive model of systemic toxicity for thousands of chemicals. • Anticipated Impact/Expected use – The ability to predict systemic toxicity using high throughput screening data would greatly enhance the Agency’s ability to prioritize chemicals. In addition, these predictive models could demonstrate the utility of high throughput assay data for informing risk decisions.
In an effort to ensure chemical safety while reducing reliance on animal testing, USEPA and L’Oréal have collaborated to address a major challenge in chemical safety assessment using alternative approaches: the prediction of points-of-departure (POD) of systemic effects. Systemic POD were curated from ToxRefDB, HESS-DB and COSMOS-DB from numerous study types totaling 4382 in vivo studies for 1201 chemicals. Observed systemic POD values in mammalian models are a complex function of chemical dynamics, kinetics, and inter- and intra-individual variability. In order to address the complexity problem, systemic POD were modeled at the study-level by leveraging study covariates (e.g., study type, strain, administration route) in addition to multiple descriptor sets, including chemical (ToxPrint, PaDEL, and Physchem), biological (ToxCast), and kinetic descriptors. Using Random Forest modeling with cross-validation and external validation procedures, study-level covariates alone accounted for approximately 20% of the variance reducing the root mean squared error (RMSE) from 0.96 log10 mg/kg/day to 0.85 log10 mg/kg/day, providing a baseline performance metric (lower expectation of model performance). A consensus model developed using a combination of study-level covariates, chemical, biological, and kinetic descriptors explained a total of 38% of the variance with an RMSE of 0.76 log10 mg/kg/day. A benchmark model (upper expectation of model performance) was also developed with an RMSE of 0.5 log10 mg/kg/day by incorporating study-level covariates and the mean POD per chemical. To achieve a representative chemical-level prediction, the minimum study-level predicted and observed POD per chemical were compared reducing the RMSE from 1.1 to 0.8 log10 mg/kg/day. Although biological descriptors did not improve model performance, the final model was enriched for biological descriptors that indicated xenobiotic metabolism gene expression, oxidative stress, and cytotoxicity, demonstrating the importance of accounting for kinetics and non-specific bioactivity in predicting systemic POD. Herein, we have generated an externally predictive model of systemic POD for use as a safety assessment tool and have generated forward predictions for thousands of chemicals