QSAR study of HIV-1 inhibitors using genetic programming and multiobjective optimization

Abstract

Variable selection is one of the most important procedures for constructing a simple and predictive model in quantitative structure-activity relationship (QSAR) studies. In this study we propose a method for variable selection using the concept of Pareto optimum and genetic programming. As a case study, QSAR model has been developed for a set of inhibitors of the human immunodeficiency virus 1 (HIV-1) reverse transcriptase, derivatives of 1-[(2-hydroxyethoxy)methyl]-6-(phenylthio)thymine (HEPT). Structural descriptors used in this study are Hansch constants for each substituent and topological descriptors. As a result of variable selection, 11 descriptors were selected from 34 original ones. Then counter-propagation (CP) neural network were constructed with selected variables. The obtained network was accurate, predictive and interpretable and the value of explanation variance was 0.875. The CP model could easily be interpreted and the result was quite similar to previous studies of QSAR models of HEPT derivatives. Moreover in order to confirm a predictive ability of the model, validation test was performed.