B-cell epitope prediction module development

Abstract

IntroductionEpitopes are categorized into two groups: T-cell and B-cell, and further separated into linear and conformational epitope. Although conformational epitope prediction is more accurate compare to sequence based epitope prediction, not as much of protein secondary structures were solved. Mapping the epitopes by experimental way is time-consuming, reversible and expensive. Even though lots of epitope predictions are available but performance is yet to be satisfied. Thus, a better sequence based epitope prediction is needed with improvement of the previous methods. ObjectiveTo improve the sequence-based B-cell epitope prediction by support vector machine (SVM). MethodsThis method is an integration of SVM and 32 physiochemical properties. The data obtained from Immune Epitope Database, IEDB, was separated into five categories: bacteria, virus, unicellular, multicellular except vertebrates, and fungus. All modules were developed by using MATLAB programming language. The accuracy of both predictors must be more than 50% (random guessing) in order to make the prediction useful. Results & DiscussionThe accuracy obtained for both predictors are higher than 50%. The benchmarking results show that the B-cell epitope prediction module has a comparable prediction results to other available predictors. ConclusionThe improved sequence based B-cell epitope prediction module could be further developed as a web-based server to benefit potential users.