Relación estructura-actividad como estrategia para la selección de moléculas candidatas para el desarrollo de inhibidores de tirosinasas

Abstract

The purpose of this work was to calibrate a Quantitative Structure-Activity Relationship (QSAR) model to predict the half maximal inhibitory concentration (IC50) of the tyrosinase activity of 581 molecules. For geometry optimization, the molecular mechanic force field (MM+) was used, followed by the PM3 semi-empirical method to refine the structures. Then, compounds were described by 5274 alvaDesc molecular descriptors and 166 MACCS structural keys, which were merged into a single dataset and subsequently reduced by the V-WSP unsupervised variable reduction. Thus, 1692 descriptors were submitted to the Genetic Algorithms (GAs) supervised variable selection process coupled with the k-nearest neighbors classifier. In this step, the simplex method was applied in order to optimize the threshold for the separation between the high and low activity classes. A model composed of eight molecular descriptors and four neighbors classifiers was retained as the optimal one (NERtrain = 0.82). This model was validated by a cross-validation protocol based on venetian blind (NERcv = 0.82) and an external test set of 174 molecules (NERtest = 0.86). In addition, the applicability domain of the model was defined, and the mechanistic interpretation of molecular descriptors was provided. The proposed QSAR model was developed using the five principles defined by the Organization for Economic Co-operation and Development (OECD) in order to guarantee its applicability.