You are here:
Deriving Points of Departure and Performance Baselines for Predictive Modeling of Systemic Toxicity using ToxRefDB (SOT)
McLaurin, K., L. Truong, G. Ouedraogo, S. Loisel-Joubert, AND M. Martin. Deriving Points of Departure and Performance Baselines for Predictive Modeling of Systemic Toxicity using ToxRefDB (SOT). Presented at SOT 2014, Phoenix, AZ, March 23 - 27, 2014.
A primary goal of computational toxicology is to generate predictive models of toxicity. An elusive target of alternative test methods and models has been the accurate prediction of systemic toxicity points of departure (PoD). We aim not only to provide a large and valuable resource of PoD, but to also scope the problem by generating floor and ceiling baseline uncertainty bounds for which to judge future models. EPA’s ToxRefDB, originally populated with pesticide registration data, has grown to incorporate guideline-like studies from the pharmaceutical industry, National Toxicology Program, and publicly available research literature. Over 6000 high quality animal studies on 1071 chemicals were captured using standardized study design, treatment and effect vocabulary. Systemic lowest and no effect levels (LEL/NEL) were obtained from each study across a diverse set of study types including systemic sub-acute (SAC), subchronic (SUB), chronic (CHR) studies as well as systemic adult effects observed in developmental (DEV) and reproductive (MGR) studies. Species and study type adjusted chemical-level NEL were derived demonstrating a floor baseline of roughly 5 orders of magnitude uncertainty (OMU; 95% CI) based on the default distribution of NEL. Using SUB to predict CHR rat and mouse to predict rat CHR NEL, ceiling baselines were established of 3 and 3.5 OMU, respectively. Further classification of study types based on exposure duration (short = SAC, DEV; medium = SUB; long = CHR), established a ceiling baseline for short vs medium, and long vs medium to be 3.3 and 3.9 OMU, respectively. Thusly, the goal of any predictive model of systemic toxicity is to improve upon the 5 OMU and approach 3-3.5 OMU, but cannot be expected to exceed the inherent uncertainty in toxicological testing and evaluation. This abstract does not necessarily reflect US EPA policy.
We aim to provide a point of departure resource for predictive models of toxicity, scoping the problem by generating floor and ceiling baseline uncertainty bounds for which to judge future models.
Record Details:Record Type: DOCUMENT (PRESENTATION/POSTER)
Organization:U.S. ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF RESEARCH AND DEVELOPMENT
NATIONAL CENTER FOR COMPUTATIONAL TOXICOLOGY