Exploring Molecular Docking And Quantitative Structure-Activity Relationships Of Oridonin Derivatives As Inhibitors Of Akt Protein Kinase

Abstract

In this work hybrid molecular docking quantitative structure activity relationship (QSAR) methodology is used to modeling and predict the inhibitory activities of some Oridonin derivatives to ward kinase B(AKT) protein. Data set consist of inhibitory activities of 57 Oridonin derivatives (as EC₅₀ in µM^-1), which can be used in treatment of type 2 leukemia. After docking of these derivative’s to B protein (AKT), the most stable structure of ligands are chosen and frozen, to calculate molecular descriptors. In the next step prescreened of descriptors are done and stepwise feature selection methods used to select the most relevel descriptors. Then the selected descriptors are used to developing multiple linear regression (MLR) and support vector machine (SVM) models. The statistical parameters of these model are; R² of 0.62 and 0.80; SE of 1.44 and 1.33 for MLR, and for SVM models R² of 0.89 and 0.82; SE of 0.44 and 0.79 for training and test sets respectively. Comparison between these valves of and other statistics reveals the superiority of SVM over MLR models. In the next step virtual screening based on the lead derivatives is outperformed to identify new efficient candidate based on ADME properties and docking studies.