Predictive QSAR Modeling of Cyclic Urea HIV-1 Protease Inhibitors Based on Linear and Non-Linear Regression Methods
Somayeh Pirhadi, Jahan Bakhsh Ghasemi · 2010
A Quantitative Structure – Activity Relationship (QSAR) analysis of cyclic urea-based Human Immunodeficiency Virus Type 1 was carried out on a set of cyclic urea-based (HIV-1) protease inhibitors. The stepwise multiple regression analysis and genetic algorithm were employed as the feature selection and model development methods. Modeling of the relationship between selected molecular descriptors and pKi data was achieved by MLR as linear and A Levenberg– Marquardt algorithm trained feed-forward artificial neural network as nonlinear methods. A comparison between the obtained results revealed the non- linear method is better than linear method. The improvement is due to the fact that the activity of the compounds demonstrates nonlinear correlations with the selected descriptors.