Estimation of rodent acute toxicity is an important task in drug design and risk assessment of chemicals. QSAR studies of toxicity of chemicals are widely used for this purpose. In this work we used the Accelrys Toxicity database, which includes information about acute rodent toxicity for 60,000 non-congeneric organic molecules. The toxicity end-points are based on LD50 (mmol/kg) values for mice and rats with four types of administration: oral, intravenous, intraperitoneal and percutaneous. We have generated a variety of QSAR models using different statistical methods: RBF-SCR, RBF NN, Random Forest, SVR; and descriptor sets: QNA (Quantitative Neighbourhoods of Atoms) descriptors, �biological� descriptors based on PASS prediction results, Dragon descriptors. The models obtained were compared using external validation sets. Differences in toxicity values observed for mice and rats were analyzed across the four types of administration. We discuss our attempts to infer the mechanism of toxicity for compounds from our models. We also report on our efforts to incorporate the most predictive ones of our models in the publicly available NCI/CADD Group web services (htts://cactus.nci.nih.gov).

QSAR modelling of rodent acute toxicity

247th National Meeting (2014)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.