A Novel Linear B-cell Epitope Prediction Method based on Position Entropy of Amino Acids

Abstract

Epitope prediction plays an important role in diagnosis, treatment of diseases and the development of antibodies. Recently, many machine learning algorithms and new strategies have been used to predict the B-Cell epitopes. However, the performance of epitope prediction is still not satisfactory. We propose the method of Linear B-cell epitope prediction base on the position entropy of amino acids and long and short-term memory (LSTM) network. We design three sets of experiments to verify the effectiveness of the model. The result of experiments indicates that the accuracy of our method can reach to 88.94%. The result also show that the position entropy of amino acids is an effective feature in B-cell epitope prediction.