The Research Of B-cell Epitope Prediction Based On The Phage Peptide Library Screening

In the course of humoral immunity, BCR can recognize the foreign antigen and generate antibody which can bind specifically with the antigen, meanwhile part of B-cell are activated to differentiate into memory B-cells, and make more rapid immune response when the pathogens invade body next time. A B-cell epitope is defined as a part of protein antigen recognized by either a particular BCR or a particular antibody molecule of the immune system.Mapping the B-cell epitope has more significance in the design of artificial vaccines, immune intervention therapy and in the terms of high-throughout antibody preparation. Currently, the most reliable way of mapping B-cell epitope is getting the 3D structure of antigen-antibody complex through crystal diffraction and NMR; however, these two experimental methods require high cost and massive manpower, also have high requirements for equipment. With some of auxiliary experiment and the increase of known epitope data, people began to predict epitope using computer. Using computer to predict the candidate epitope at first and through biological experiments to validate following. To predict the epitope with the combination of experimental and computer methods can both guarantee the accurate results and save cost.The B-cell epitope prediction based on phage peptide library screening is one of the experimental and computer combined methods. Firstly, the method need high affinity mimotope sequences from phage peptide library screening experiment, then search the amimo acids in the surface of antigen which match the mimotope sequence through the computer, and determine the candidate epitope. In recent years, with the growing of mimotope sequence and the 3D structure of antigen-antibody complex, many B-cell epitope prediction methods based on phage peptide library screening have been proposed, and showed better prediction performance in a few test cases. However, there is no benchmark dataset for B-cell epitope prediction based on peptide phage library screening so far. In addition, a complete evaluation system for comparison and analysis the performance of the prediction algorithm is badly needed.This research works at construction of benchmark dataset for B-cell epitope prediction based on phage peptide library screening, establishment of the evaluation system for algorithms, giving a more sensitive B-cell epitope prediction method based on the structure features of antigen protein and peptide phage library screening.Firstly, the paper makes a comprehensive study of the existing B cell epitope prediction methods based on phage peptide library screening, integrates the relational information of MimoDB, PDB, CED and IEDB databases, constructs a common benchmark dataset. Using the benchmark dataset and a representative dataset, the paper tests five published B-cell epitope prediction methods based on phage peptide library screening: Mapitope, PepSurf, Pepitope, Pep-3D-Search and EpiSearch. These five methods either provide source code or provide web service freely. Through the benchmark dataset and using sensitivity, specificity, precision, and MCC evaluation parameters, the paper establishes a comprehensive evaluation system for B cell epitope prediction methods based on peptide phage library screening, and makes a complete performance evaluation of the five methods.Based on the comprehensive evaluation,this paper proposes a more sensitive B-cell epitope prediction method based on the structure features of antigen protein and phage peptide library screening. The method frirstly classify the amino acids of antigen protein based on structure features and support vector machine. Then map the mimotopes to the antigen protein. The searching algorithm adopts the idea of partition based on the current study, divides the antigen surface into lots of overlapping patch regions to predict the epitope. In the step of constructing undirected graph for every patch, the algorithm initially tries to use a variable distance threshold to define whether two vertices have an edge. In addition, the paper first adopts a comprehensive search method to ensure obtaining the optimal solution.Finally, the paper implements the newly proposed algorithm. Through comparing with the other five methods, the sensitivity of the new method has a significantly improvement. The study not only has a great significance for theory research of B cell epitope prediction, but also pushes its development to application.