Updated version at IIIT Delhi

Artificial neural network based B-cell epitope prediction server

B-cell and epitope of antigen

Recurrent neural network

The aim of ABCpred server is to predict B cell epitope(s) in an antigen sequence, using artificial neural network. This is the first server developed based on recurrent neural network (machine based technique) using fixed length patterns.
Algorithm: The machine-learning technique need fixed length patterns for training or testing whereas B-cell epitopes vary 5 to 30 residues as reported in literature(Bcipep database Bcipep). In order to overcome this problem we made an attempt to create datasets of fixed length patterns from B-cell epitopes by eliminating or adding residues at terminals. The dataset used for training and testing consists of 700 B-cell epitopes and 700 non B-cell epitopes (random peptides) of maximum length of 20 residues. We tried different neural networks and achived an accuracy of 65.93% using recurrent neural network.
Users can select window length of 10, 12, 14, 16 and 20 as predicted epitope length. It presents the results in graphical and tabular frame. In case of graphical frame, this server plot the epitopes in blue color along protein backbone (black color), which assist the users in rapid visulaziation of B-cell epitope on protein. The tabular output is in the form of a table, which will provide the aminoacids length from N-terminal to C-terminal in a protein predicted by the server.
The server is able to predict epitopes with 65.93% accuracy using recurrent neural network.
Please cite following paper if you are using ABCpred server
Saha, S and Raghava G.P.S. (2006) Prediction of Continuous B-cell Epitopes in an Antigen Using Recurrent Neural Network. Proteins,65(1),40-48 PMID: 16894596
Contact :  G.P.S. Raghava
Bioinformatics Centre
Institute of Microbial Technology,India