[TEpredict: software for T-cell epitope prediction]

A program named TEpredict was developed for T-cell epitope prediction. Original models for T-cell epitope prediction were constructed by means of Partial Least Squares regression method on the basis of data, extracted from the IEDB (Immune Epitope Database)--the most complete resource of experimental peptide-MHC binding data known to date. TEpredict is also able to predict proteasomal processing of protein antigens, and the ability of produced oligopeptides to bind to TAP (Tansporters Associated with Processing). TEpredict could exclude peptides, shearing local similarity with human proteins, from the set of predicted T-cell epitopes. It is also able to estimate expected population coverage by selected peptides, using known HLA allele genotypic frequencies data. The majority of produced models demonstrated high sensitivity of predictions (0.50-0.80) concurrent with high specificity (0.75-0.99). TEpredict was shown to be highly competitive or even superior in comparison with such programs as ProPred1, SVRMHC, SVMHC and SYFPEITHI. TEpredict demonstrated high quality of predictions and we expect that it could become a useful tool in the development ofpolyepitope vaccines against dangerous human pathogens, including HIV, influenza etc. The program and its source code could be freely downloaded from the project web-site: http://tepredict.sourceforge.net.