B-cell Epitope Prediction Method Based On Sequences And Deep Ensemble Architecture

Epitope is the basic unit that stimulates the organism's immune system to produce specific immune response during the interaction between antigen and organism.B-cell epitope is known as some regions on the surface of a pathogenic antigen which can be recognized by an antibody or B-cell receptor(BCR),these certain regions on the surface of antigen induce human immune response.Identification of B-cell epitope is to aid the design of molecules that can mimic the structure and function of a genuine epitope and replace it in antibody diagnostics and therapeutics,as well as the design of the potentially safer vaccine.Accurate prediction of B-cell epitope is not only helpful for basic immunological research,but also for disease prevention and diagnosis.The most reliable methods for identification of an B-cell epitope is experimental methods,such as X-ray crystallography and NMR techniques,but they are time consuming and expensive.Candidate epitopes that are selected by computational methods which prior to laboratory experiments can lead to both significantly reducing the experimental cost and substantially accelerating the work efficiency.At present,according to the input data and calculation strategy,the B-cell epitope prediction methods based on calculation can be divided into three categories: structure-based prediction,mimotope-based prediction and sequence-based prediction.Structure-based prediction methods employ the epitope related features such as geometric properties and physical and chemical properties,and then predict the epitopes by scoring or machine learning methods.Mimotope-based prediction methods employ the mimotope sequences which were obtained from phage display experiments and the 3D structure of antigen as input.Sequence-based prediction methods use the antigen sequence only.They extract the epitope related features of the antigen sequence,form the feature vector or feature matrix,and then predict the epitopes by scoring or machine learning.Due to the development of sequencing technology,antigen sequence are more easily obtained.In recent years,several sequence-based B-cell epitope prediction methods have been proposed,and these methods have achieved better prediction results on some datasets.In this work,we construct a deep ensemble architecture for B-cell epitope prediction based on antigen sequences.We adopted one hot vector coding and physico-chemical properties schemes for encoding protein sequence fragments,and constructed seven independent convolutional neural network respectively,then the weighted average method was used to integrate the seven networks.The proposed method is evaluated on the testing datasets of BepiPred 2.0.The experimental results show that the method achieves an AUC of 0.771,a sensitivity of 0.711,and a Matthews correlation coefficient of 0.222.In addition,we also evaluate the performance of our method on 13 independent testing cases and the results are superior to the existing methods.Therefore,our proposed method based on protein sequence and deep ensemble architecture can effectively predict B-cell epitopes.