QSAR Model Based Gradient Boosting Regression of N-Arylsulfonyl-Indole-2-Carboxamide Derivatives as Inhibitors for Fructose-1,6-Bisphosphatase

Abstract

Background: Due to the complication caused by conventional drugs, global attention has been focused on the development of novel drugs. As a consequence, a potential theory to put T2DM under control is of great medical significance. Methods: we used the heuristic method to establish the linear model and used Gradient Boosting Regression to establish the nonlinear model of Fructose-1,6-Bisphosphatse inhibitor successively. In this study, 84 derivatives of N-Arylsulfonyl-Indole-2-Carboxamide were introduced into the models, and two outstanding QSAR models with 2 molecule descriptors were established successfully. Results: Gradient Boosting Regression rendered a good correlation with R2 of 0.943 and MSE of 0.135 for the training set, 0.916 and 0.213 for the test set, which also proves the feasibility of the implementation of the new method GBR in the field of QSAR. Meanwhile, the optimal model displayed wonderful statistical significance. Conclusion: This study makes an attempt at the application of a new method of GBR in QSAR and proves GBR as a promising tool for further study of CADD.