Antioxidant activity of NSAIDs-Se derivatives: predictive QSAR-machine learning models

Abstract

The study employed the random forest (RF) and artificial neural network (ANN) methods based on the quantitative structure–activity relationship (QSAR) techniques to analyze NSAIDs-Se derivatives and their antioxidant abilities. The best predictive models by the RF method yielded a coefficient of determination (R²) value of 0.868 for the training set, and the root mean square error (RMSE) of the test set was only 0.053. For the QSAR-ANN, the best predictive models resulted in an R² value of 0.935, and the RMSE of the test set was 0.068. Based on the best models, the steric, electrostatic, and enthalpy descriptors are found to be related to antioxidant prediction. Thus, to extend the predictive ability of the obtained QSAR-ML models, an external set was collected from a later publication of NSAIDs-Se derivatives with experimental antioxidant abilities. The efficacy of two QSAR models in forecasting the antioxidant abilities of an external set of NSAIDs-Se derivatives was evaluated. The QSAR with machine learning models demonstrated high efficiency in predicting the antioxidant abilities of the external NSAIDs-Se set with an RMSE of the external set in the range of 0.074–0.087. Therefore, the results suggest that fine-tuning machine learning-based QSAR studies can aid in the design of novel NSAIDs-Se derivatives with highly efficient antioxidant prediction.