BayesB (version 1.0) is a web
server for predicting linear B-cell epitopes on protein sequence. The server employs the use of Bayes Feature Extraction for encoding the feature vectors in the support vector machines algorithm. The best prediction model (on 20-mer window) is shown to achieve an overall accuracy of 73% and area under the ROC curve (AROC) of 0.80. The SVM model was trained and tested using a dataset of 1402 sequences (training: 601 epitopes and 601 non-epitopes, testing: 100 epitopes and 100 non-epitopes).
B-cell epitopes are regions on antigenic proteins recognised by the variable domains on B-cell antibodies. The prediction of B-cell epitopes is important for assisting researchers in the design and development of diagnostic tests, vaccines and therapeutic proteins (learn
more about B-cell epitopes).
Left: Structure of a pre-B cell receptor (PDB:2H32).
Right: Structure of Influenza virus hemagglutinin complexed with a neutralizing antibody (PDB:1E08).
Please cite the following article for results obtained
from this server:
Lawrence JK Wee, Jason YW Kam, Diane Simarmata, Lisa FP Ng and Victor JC Tong. BayesB: Server for SVM Prediction of Linear B-cell Epitopes using Bayes Feature Extraction. (Submitted).
Server was lasted updated on 19 Jan 2010.