Selection of nutraceutical compounds as COX inhibitors by molecular topology

Abstract

QSAR based on molecular topology (MT) has proven to be a very efficient method in drug design and discovery. In this study, some models based on MT have been obtained by linear discriminant analysis (LDA) and artificial neural networks (ANN). Later on, the models were applied to the search of new cyclooxygenase (COX) inhibitors showing anti-inflammatory activity. Moreover, an external validation test has been carried out, yielding 80 % of correct classification within the active compounds and 78.6 % within the inactive. The results from ANN showed a correct classification percentage above 85 % for the test set and of 90 % for the external validation set. The accuracy of the models was also checked using the literature data, upon the carrageenan-induced mice paw edema test. In this case, the models were capable to classify correctly four out of five active compounds as well as two out of the two inactive ones, which enabled the models’ optimization. Finally, a virtual screening on a nutraceutical database was performed, from which ten compounds were selected for their potential COX inhibitory activity. The results shown here enhance MT’s role as a very efficient tool for the discovery of new COX inhibitors with potential anti-inflammatory activity.