Predicting important classes of chemokine family based on kernel method Original Research Article

Chemokines are a family of chemotactic cytokines that have important physiological roles in a wide range of disease processes. Identifying novel class of chemokines by kernel methods can provide insights for the functional studies of the human uncharacterized proteins. In this study, a support vector machine learning system was trained to predict two main classes of chemokines (CC and CXC classes) based solely on amino acid composition and associated physicochemical properties. Further, the effect of different kernel functions and learning methods were investigated. The cross-validation results demonstrated that this kernel method performed well in identifying two main classes of chemokines.