An ensemble classifier based prediction of G-protein-coupled receptor classes in low homology

Abstract

G-protein-coupled receptors (GPCRs) play an important role in physiological processes which are the targets of more than 50% of marketed drugs. In this research, we use a hybrid approach of predicted secondary structural features (PSSF) and approximate entropy (ApEn) as the feature selection method for predicting G-protein-coupled receptors in low homology. The low homology dataset is used to validate the proposed method for its objectivity. The classification model based on the fuzzy K-nearest neighbor classifier has been utilized on the classification of membrane proteins data. In order to enhance the prediction accuracies, here we propose an ensemble classifier as the prediction engine. Compared with the previous best-performing method, the success rate is encouraging. The reliable results also demonstrate the proposed method could contribute more to the characterization of various proteomes and further utilized in neuroscience.