Prediction of Antigen Epitopes on Protein Surfaces Based on Support Vector Machine

Abstract

B-cell epitope prediction is important for vaccine design, development of diagnostic reagents and for studies to elucidate the interactions between antigen and antibody on a molecular level. Here, we present a new epitope prediction method based on six different scoring functions and exploited LibSVM to predict the antigenic epitopes in protein surface. Using bound structures of the testing dataset, the method was able to predict antigenic epitopes with 50.6% sensitivity, 62.9% specificity, 19% precision and an AUC value of 0.616. While using unbounded structures of the testing dataset, the performance of the method was nearly the same. Compared with another epitope prediction method EPCES, the performance of the method is statistically similar. The results suggest that more effective features that discriminate epitopes from non-epitopes may further improve the performance of the prediction method. Also, the new algorithms for predicting the epitopes are desired and the construction of large with non-redundant datasets is strongly needed.