Predicting cleavage sites in exogenous antigen using weighted SVM

In the pathway of helper T-cell immunity, the exogenous antigen is firstly degraded into short peptides, and then some peptides can be bond to certain major histocompatibility class II (MHC-II) molecule to develop peptide-MHC complex, finally immune response is activated once receptors on helper T-cell recognize the complex. These binding peptides is named as `helper T-cell epitopes´, and they can be useful for the epitope-based vaccine design. Proteases play an important in MHC-II ligand generation and thus in the regulation of specific immune responses, therefore modeling the process of protease cleavage is the preliminary step for epitope recognition. To our knowledge, this work is the first effort for the prediction of cleavage sites in exogenous antigens involved in MHC-II antigen presentation. In this paper, we collect the peptides sequences from IEDB database, and then obtain their parent exogenous antigens from Uniprot database. We use sliding windows to move along the antigens, and transform original task into a problem of unbalanced data classification. The weighted support vector machine is used to construct the prediction models, and computational experiments demonstrate the satisfying results in terms of both independent test and cross-validation.