Artificial neural networks incorporating chirality codes have been shown to predict mainly non-enzymatic reactions involving enantiomeric compounds.  Herein, we identify the most optimal artificial neural networks employing chirality codes to predict enzymatic catalytic parameters (Vmax, Km, and Vmax/Km) for CYP2C19 toward a diverse array of enantiomeric substrates.    Ab initio methods using Gaussian software are employed to determine the physiochemical properties and optimized three-dimensional structure for pairs of enantiomeric substrates.  This information is used to generate chirality codes for compounds that serve as inputs to a back propagation neural network.  Different architectures for the neural networks are analyzed by leave-one-out cross correlations to identify the best network.  While studies with Vmax and Km are on-going, we achieved a network predicting Vmax/Km with a leave-one-out cross-correlation r2 0.88 and slope 0.89.  Importantly, we are further assessing the utility of neural networks by comparing predictions to actual parameters for previously uncharacterized CYP2C19 reactions.

Predicting catalytic parameters of enantioselective oxidations by CYP2C19 using a chirality code and artificial neural networks

243rd National Meeting (2012)

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