Knowledge of drug metabolism is critical to understanding bioavailability, prodrug design, and metabolic toxicities.  Cytochrome P450s are responsible for the metabolism of most drugs.  We have developed in silico models using Artificial Neural Network Ensembles (ANNEs) for ultra-high-throughput prediction of likely sites of oxidation and resulting metabolites for multiple CYP isoforms including 1A2, 2A6, 2B6, 2C8, 2C9, 2C19, 2D6, 2E1, and 3A4.  These models employ highly accurate atomic descriptors accounting for charge, reactivity, steric effects, and other aspects of an atom's local environment with no requirements for 3D structure of either substrate or enzyme.  Influence of specific descriptors and their effect on model performance, as well as importance of data curation in model building, will be discussed.  Performance of the models will be shown across hundreds of training and test set molecules, along with specific examples where apparently errant predictions were later borne out by subsequent experiments

Accurate and superfast metabolite prediction: Have your cake and eat it, too

243rd National Meeting (2012)

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