Use of Artificial Neural Networks in a QSAR study of biological activities of derivatives of Benzimidazoles on CoII-Loaded Escherichia coli Methionine Aminopeptidase

The paper describes quantitative structure activity relationship (QSAR) study of IC50 values of derivatives of Benzimidazoles on CoII-Loaded Escherichia coli Methionine Aminopeptidase. The activity of a set of 32 inhibitors was estimated by means of multiple linear regression (MLR) and artificial neural network (ANN) techniques. The results obtained using the MLR method indicate that the activity of derivatives of benzimidazole on CoII-Loaded Escherichia Coli Methionine Aminopeptidase depends on different parameters containing topological descriptors, Burden eigen values , 3D Morse and 2D autocorrelation descriptors. Then, these descriptors was used as inputs for an artificial neural network model. The best artificial neural network model was a fully-connected, feed forward back propagation network with a 5-4-1 architecture. Standard error for the training set using this network was 0.193 with r = 0.996 and standard error for the prediction set was 1.41 with r = 0.802. Also, a leave six out cross-validation method have been used for different sets of compounds. comparison of the results indicates that the ANN method has a better predictive power than the MLR method for prediction of biological activity of derivatives of Benzimidazoles.