Comprehensive Modeling and Molecular Docking of Phosphodiesterase Inhibitors
Abstract In the present study, the biological activity of some pharmaceutical molecules was investigated and predicted by using quantitative structure-activity relationship (QSAR) studies and molecular docking. The aim of this study is to apply QSAR and molecular docking methods to predict and calculate the half maximal inhibitory concentration (IC50) of phosphodiesterase (PDE) 10A. To apply QSAR method at first multiple linear regression (MLR), genetic algorithm (GA), and successive projection algorithm (SPA) were used to select the best descriptors related to PDE 10A inhibitory activity. The selected descriptors were then applied as inputs to construct MLR and a non-linear support vector machine (SVM). Also we used molecular docking method to extract the descriptors and then by using MLR and SVM methods a linear and non-linear models developed respectively. Consequently, a comparison of the results of QSAR and docking study indicated that the non-linear SVM-SPA2 in QSAR study and SVM-MLR in molecular docking study had a much better prediction power than the other models. Finally, the Y-scrambling and cross-validation tests were used to evaluate the validity of the obtained model and the results of these tests indicates that the model is appropriate for using in prediction.