QSAR analysis of cyclooxygenase inhibitors selectivity index (COX1/COX2): Application of SVM-RBF and MLR methods

Abstract

Background: Anti-inflammatory inhibitors of cyclooxygenase 2 (COX2) have been shown to increase the risk of adverse cardiovascular events in clinical trials. The studies showed that such adverse events could be due to COX2-induced suppression of prostaglandin I2 (PGI2) synthesis. These adverse effects related to the degree of COX2 selectivity of NSAIDs. Study of the selectivity index of COX1/COX2 is important for development of the new NSAID drugs. Prediction methods of such index have been interested by scientists. Methods: The selectivity index of a number of 68 molecules from 8 different chemical groups was predicted using MLR and SVM-RBF models. Calculated structural and physicochemical parameters, using the energy optimized molecular structures were applied to develop the desired models. The developed models were validated using LMO, external test set and Y-randomization methods. Results: Regression coefficient of developed MLR model was 0.825 and 0.752 for training and test sets, while for SVM-RBF model it was 0.628 and 0.863 for training and test sets. The RMSE of the developed models were 0.08, 0.06, 0.29 and 0.16 respectively for train (MLR, SVM-RBF) and test (MLR, SVM-RBF) datasets. Conclusion: The validation results showed the higher prediction capability for SVM-RBF in comparison with MLR models. The selected descriptors showed the contribution of electronic parameters in conjunction with size and shape parameters in selectivity of studied compounds. é 2015 The Authors.