Prediction of Linear B-Cell Epitopes Using AAT Scale

Abstract

The prediction of B-cell epitopes is of great importance for computer acid vaccine design and immunodiagnostic test. Although it is said that a large majority of B-cell epitopes are conformational, experimental epitope identification has focused primarily on linear B-cell epitopes. A number of computational methods have been developed for the prediction of linear B-cell epitopes, but few of them can give us a convincible result. In this paper, a new method, call AAT-fs is proposed which focus on the amino acid triplet (AAT) antignenicity scale. After using AAT scale to create input vectors, we develop a support vector machine (SVM) for the classification which is trained utilizing RBF kernel on homology reduced datasets with fivefold cross- validation. The AAT-fs method gets the better performance than AAP scale, BCPred and other existing B-cell epitope prediction algorithms. It can be expect that with the rapid development of B-cell epitope identification experimental technology, the dataset will increase and AAT-fs can achieve better result.