B-Cell Linear Epitope Prediction Using Transformer Encoder

Youjin Kim, Jinhee Park, Junseok Kwon

Abstract: When a pathogen invade a host, the antibody binds to a specific part of the pathogen's antigen and neutralizes the pathogen, thereby performing immune activities. The site of the antigen recognizable by the antibody is called epitope. Accurate prediction of epitope is very important for vaccine design, targeted therapy and understanding of immune system. Since the assay experiment to determine the epitope of the antigen is usually an extensive laboratory work, serveral studies have been conducted to predict the B-Cell epitope quickly and conveniently by using machine learning. In this paper, we propose a novel deep neural network based on transformer encoder and 1-dimensional ResNet to predict B-cell linear epitopes.