Sequence-based B-cell epitope prediction by using associations in antibody-antigen structural complexes

B-cell secreted antibodies play a critical role in fighting against the invaders and abnormal self tissues. Identifying the epitope on antigens recognized by the paratope on antibodies can enlighten the understanding of this important immune mechanism. Predicting B-cell epitope can also pave the way for vaccine design and disease therapy. However, due to the high complexity of this problem, previous prediction methods that focus on linear and conformational epitope are both unsatisfactory. In this work, we propose a novel method to predict B-cell epitopes, when a pair of sequences is given, by using associations and cooperativity patterns from a relatively small antigen-antibody structural data set. More exactly, our classifier is trained on only PDB protein complexes, but it can be applied to any sequence data. Our evaluation results show that the accuracy of our method is very competitive to, sometimes even much better than, previous structure-based prediction methods which have a smaller applicability scope than ours.