A QSAR study of antiplatelet agents using artificial neural network - correlation with micelle-water partition coefficient

Antiplatelet activity ex vivo data reported for 2-substituted phenyl- and benzimidazolyl-5-methyl-4-(3-pyridyl) imidazoles have been analysed using BP type ANN. Using micelle-water partition coefficient as an independent descriptor, a network system (1-3-1) produced very good duplication of observed activities (r=0.860, SD=0.183, N=21) in the training cycle. The results provide an improved model for prediction of antiplatelet activity.