In silico identification of human pregnane X receptor activators from molecular descriptors by machine learning approaches

In the current study, computational models for hPXR activators and hPXR non-activators were developed using support vector machine (SVM), k-nearest neighbor (k-NN), and artificial neural networks (ANN) algorithms. 73 molecular descriptors used for hPXR activator and hPXR non-activator prediction were selected from a pool of 548 descriptors by using a multi-step hybrid feature selection method combining Fischer's score and Monte Carlo simulated annealing method. The y-scrambling method was used to test if there is a chance correlation in the developed SVM model. In the meantime, five-fold cross validation of these machine learning methods results in the prediction accuracies of 87.2-92.5% for hPXR activators and 73.8-87.8% for hPXR non-activators, and the prediction accuracies for external test set are 93.8-95.8% for hPXR activators and 86.7-92.8% for hPXR non-activators. Our study suggested that the tested machine learning methods are potentially useful for hPXR activators identification.