Affiliation: Gillings School of Global Public Health, Department of Environmental Sciences and Engineering
Traditional reproductive toxicity testing is inefficient, animal intensive and expensive with under a thousand chemicals ever tested among the tens of thousands of chemicals in our environment. Screening hundreds of chemicals through hundreds high-throughput biological assays generated a validated model predictive of rodent reproductive toxicity with potential application toward large-scale chemical testing prioritization and chemical testing decision-making. Chemical classification for model development began with the uniform capturing of the available animal reproductive toxicity test information utilizing an originally developed relational database and reproductive toxicity ontology. Similarly, quantitative high-throughput screening data were consistently processed, analyzed and stored in a relational database with gene and pathway mapping information. Chemicals with high quality in vivo and in vitro data comprised the training, test, external and forward validation chemical sets used to develop and assess the predictive model based on eight selected features generally targeting known modes of reproductive toxicity action. In three case studies, the forward validated predictive model reduced the overall costs of reproductive toxicity testing by roughly twenty percent. The model provides a starting point for the future of reproductive toxicity testing.