Novel alignment method of small molecules using Hopfield Neural Network

Abstract

In the methodologies of 3D-QSAR such as Comparative Molecule Field Analysis (CoMFA), appropriate molecular alignment is definitely required for predictive data modeling and correct analysis of the model. In this study, the novel method for molecular alignment using Hopfield neural networks is proposed. A Hopfield neural network is one of artificial neural networks that developed by Hopfield and Tank in 1985, and it has been mainly used as Content-Addressable Memory. Recently, ability of Hopfield neural network to perform combinatorial optimization was reported by Feuilleaubois et al. and several applications were presented. In our method, each molecule is represented by four chemical properties (Hydrophobic, Hydrogen-bonding donor, Hydrogen-bonding acceptor and Hydrogen-bonding donor/acceptor), and then the properties among molecules are corresponded by HNN. In order to validate usefulness of this method, 3D-QSAR analyses of cyclooxygenase-2 (COX-2) inhibitors were performed. All compounds in the data set were superposed by our method, and then QSAR model using CoMFA field variables was calculated by Partial Least Squares (PLS) method. As a result, the robust regression model (R^2=0.93, Q^2=0.70) with five components was obtained by leave-one-out procedure, and it could give meaningful counter map.