Computer-aided molecular design with BP-ANN and global optimization algorithm

Abstract

Abstract A new method of computer-aided molecular design (CAMD), in which back-propagation artificial neural network (BP-ANN) is employed to establish the quantitative structure-activity relationship (QSAR) model and the α BB algorithm is applied to search for the greatest activity and the corresponding structural parameters, is presented in this paper. In order to achieve precise QSAR models with BP-ANN, a self-clustering algorithm and a genetic algorithm are introduced to the modeling process. The convex lower bounding function for BP-ANN model is constructed to perform the α BB algorithm. The efficiency of the proposed method is estimated by an illustrative example, and the limitation of the method is discussed.