QSAR study of prolylcarboxypeptidase inhibitors by genetic algorithm: Multiple linear regressions

Abstract

The predictive analysis based on quantitative structure activity relationships (QSAR) on benzimidazolepyrrolidinyl amides as prolylcarboxypeptidase (PrCP) inhibitors was performed. Molecules were represented by chemical descriptors that encode constitutional, topological, geometrical, and electronic structure features. The hierarchical clustering method was used to classify the dataset into training and test subsets. The important descriptors were selected with the aid of the genetic algorithm method. The QSAR model was constructed, using the multiple linear regressions (MLR), and its robustness and predictability were verified by internal and external cross-validation methods. Furthermore, the calculation of the domain of applicability defines the area of reliable predictions. The root mean square errors (RMSE) of the training set and the test set for GA-MLR model were calculated to be 0.176, 0.279 and the correlation coefficients (R 2) were obtained to be 0.839, 0.923, respectively. The proposed model has good stability, robustness and predictability when verified by internal and external validation. The predictive model using quantitative structure-activity relationships (QSAR) on benzimidazolepyrrolidinyl amides as prolylcarboxypeptidase (PrCP) inhibitors was derived based on internally and externally validated robust GA-MLR model.