Abstract

The identification of major histocompatibility complex (MHC) class-II restricted peptides is an important goal in human immunological research leading to peptide based vaccine designing. These MHC class II peptides are predominantly recognized by CD4+ T-helper cells, which when turned on, have profound immune regulatory effects. Thus, prediction of such MHC class-II binding peptide is very helpful towards epitope based vaccine designing. HLA-DR proteins were found to be associated with autoimmune diseases e.g. HLA-DRB1*0401 with rheumatoid arthritis. It is important for the treatment of autoimmune diseases to determine, which peptides bind to MHC class II molecules. The experimental methods for identification of these peptides are both time consuming and cost intensive. Therefore, computational methods have been found helpful in classifying these peptides as binders or non-binders. We have applied negative selection algorithm, an artificial immune system approach to predict MHC class-II binders and non-binders. For the evaluation of the NSA algorithm, five fold cross validation has been used and six MHC class-II alleles have been taken. The average area under ROC curve for HLA-DRB1*0301, DRB1*0401, DRB1*0701, DRB1*1101, DRB1*1501, DRB1*1301 have been found to be 0.75, 0.77, 0.71, 0.72, 0.69, and 0.84, respectively indicating good predictive performance for the small training set.

Key words: Negative selection algorithm, MHC class-II peptides, artificial immune system, epitope, vaccine designing, human immunology.