期刊名称：Journal of Materials and Environmental Science
印刷版ISSN：2028-2508
出版年度：2017
卷号：8
期号：5
页码：1532-1545
出版社：University of Mohammed Premier Oujda
摘要：Diabetes mellitus is one of the most common chronic metabolic diseases, characterized by chronic hyperglycaemia and the development of diabetes-specific microvascular and macrovascular pathology. Prolonged hyperglycemia is a primary causal factor of several diabetic complications. The flavone (phenyl-benzopyrane) and its derivatives are potent inhibitors agents, these compounds inhibit Aldose Reductase (ALR2) enzyme. A study of quantitative structure-activity relationship (QSAR) is applied to a set of 29 molecules derived from phenyl-benzopyrane, in order to predict the ALR2 inhibitory biological activity of the test compounds and find a correlation between the different physic-chemical parameters (descriptors) of these compounds and its biological activity, using principal components analysis (PCA), multiple linear regression (MLR), multiple non-linear regression (MNLR) and the artificial neural network (ANN). We accordingly propose a quantitative model (non-linear and linear QSAR models), and we interpret the activity of the compounds relying on the multivariate statistical analysis. The topological descriptors were computed, respectively, with ACD/ChemSketch and ChemBioOffice 14.0 programs. A good correlation was found between the experimental activity and those obtained by MLR and MNLR respectively such as (R = 0,80 and R2 = 0,64) and (R = 0,83 and R2 = 0,69), this result could be improved with ANN such as (R = 0,88 and R2 = 0,77) with an architecture ANN (5-2-1). To test the performance of the neural network and the validity of our choice of descriptors selected by MLR and trained by MNLR and ANN, we used cross-validation method (CV) such as (R = 0,833 and R2 = 0,693) with the procedure leave-one-out (LOO). This study show that the MLR and MNLR have served to predict activities, but when compared with the results given by an 5-2-1 ANN model we realized that the predictions fulfilled by this latter was more effective and much better than other models. The statistical results indicate that this model is statistically significant and shows very good stability towards data variation in leave-one-out (LOO) cross validation.
