A Support Vector Machine-Based Method for Predicting Chemokine Receptors Types

Chemokine receptors represent a prime target for the development of novel therapeutic strategies in a variety of disease processes. The prediction of interesting proteins types by computational methods can provide new clues in functional studies of uncharacterized proteins without performing extensive experiments. Support vector machine (SVM) is a new kind of approach to supervised pattern classification that has been successfully applied to a wide range of computational biology fields. In this study, a SVM classifier was implemented to predict two main types of chemokine receptors based solely on amino acid composition and associated physicochemical properties. The performance on the tree kernel method we developed is comparable to that of other kernels while giving distinct advantages when evaluated through 10-fold cross-validation technique, indicating the current approach may serve as a useful tool for further investigating the processes of cell molecular mechanism of this important family. The experimental results also show that the features and the classifiers in detecting chemokine receptors types are effective.