Artificial Immune System Algorithm for Peptide-based Vaccine Design

Peptide based vaccines play an important role in activating the immune response. The small peptides derived from target proteins (epitopes) of an invading pathogen, bacteria, or virus bind with the MHC molecules are recognized by CD4+ T cells or CD8+ T cells. MHC class I molecules are recognized by CD4+ T cells and MHC class – II molecules are recognized by CD8+ T cells. It is very time consuming and expensive to predict those small proteins i.e. peptides, which will bind to MHC molecules for immune response in the laboratory from protein sequence. Therefore, various computational learning algorithms have been used to identify the binders and non-binders. However, the number of known binders / non-binders (BNB) to a specific MHC molecule is limited in many cases, which, still is a computational challenge for binder/ non-binder prediction. In order to improve predictions using a learning algorithm the training data sets has to be sufficient. Here, an artificial immune system approach for small set of known binding and non-binding peptides has been used. In present study, which is an extension to our earlier work for MHC Class-II alleles, we have taken the MHC class I alleles for Sars Corona virus (Tor2 Replicase polyprotein 1ab) for HLA-A and HLA-B type molecules. Five different MHC class I alleles 3 for HLA-A type molecules and 2 for HLA-B type molecules with 5 fold cross validation have been retrieved from IEDB database for BNB. The average area under ROC curve have been found to be 0.70 to 0.81 for various alleles varies small training sets and in some cases it is 0.92.