T cell epitope prediction has always been an important research topic in immunotherapy and vaccine research. Previously Dr. Darren Flower and Dr. Irini Doytchinova developed a computational algorithm  named the additive method to predict T cell epitopes. The additive method is derived from Free-Wilson approach, which assumes that each substituent makes an additive and independent contribution to the biological activity. The additive method considers three types of interactions: the interaction between individual amino acids and the binding site, the interaction between adjacent and every second amino acids and their effects on binding. A number of allele specific models have been developed in Dr. Darren Flower's group since, including both human and mouse models. The additive method has been implemented into a web server named MHCPred, which can be used for online T cell epitope prediction.

The web server EpiJen is a further development of the additive approach. There are many factors in epitope recognition and not every peptide with high affinity to MHC proteins are epitopes. We try to model several important stage of the MHC degradation pathway: proteasome cleavage, TAP binding and MHC binding. The process is summerised in figure 1.

Figure 1. A flow chart of EpiJen prediction process.

The result output from EpiJen often consists of a very small number of peptides, eg. 5%. We believe that the predictivity of EpiJen is better than other T cell epitope prediction algorithms as it also considers proteasome cleavage and TAP binding, therefore it is able to mimic the MHC binding machnism in real life. At present only HLA class I alleles are included in the prediction. We aim to include a wider range of models as research progresses.


Doytchinova, I. A., P. Guan, D. R. Flower. EpiJen: a server for multi-step T cell epitope prediction. BMC Bioinformatics, 2006, 7, 131.

[ Edward Jenner Institute] [ Drug Design Group] [ Links] [ Contact Us]
  Home Page
  © The Edward Jenner Institute for Vaccine Research 1999