The SVM Method Applying In The T Cell Epitopes Prediction

In T-cell mediated specific immune response, the T-cell receptor (TCR) only recognizes the peptide binding to major histocompatibility complex (MHC) molecular. The formation of the peptide-MHC complex depends on the antigen processing and presentation pathway. Endogenous antigens (e.g. viruses, tumor antigens) need to be degradated by proteaome, transported by transporter associated with antigen processing (TAP) and bound by MHC classâ… molecule before recognized by cytotoxic T lymphocytes (CTL), correspondingly, peptides generated from this pathway are called CTL eptitopes; Exogenous antigens (e.g. toxins produced by bacterias) also need to be degradated by lysosomal enzyme and bound by MHC classâ…¡molecule before recognized by helper T cell (Th), and peptides generated from this pathway are called Th eptitope. Gernelly speaking, the antigen processing and presentation pathway determines the selection of T cell to eptitope. In order to further study the biological mechanism of the processing and presentation of antigen, and improve accuracy and rationality of T cells epitope prediction, support vector machine (SVM) was used to theoretically study following three important selective stages in the antigen processing and presentation pathway.1. The ubiquitin-proteasome system of the eukaryote plays an importance role in the endogenous antigen processing and presentation pathway. In order to further study the specificity of the proteasome cleavage sites, the support vector classifier (SVC) was used to build the predictive model of proteasomal cleavage sites and the predictive accuracy of the model is 83.1%. Compared to other models with the same test set, the performance of this model is more satisfying, The specificities of the cleavage sites and their adjacent positions come from analysis based on the weight coefficient of the amino acids to cleavage sites in the predictive model, showing the information about interaction of the proteasome with an antigen protein, which demonstrates that the proteasome cleaves the target protein selectively, but not randomly. This study is helpful to further reveal intrinsic mechanism how proteasome cleave antigen protein.2. In the endogenous antigen processing and presentation pathway, MHC classâ… molecules play a critical role in initiating and regulating immune responses. Peptide must be bound to an MHC classâ… molecule before recognized by the cytotoxic T lymphocytes (CTL), but only certain peptides can bind to any given MHC class I molecule. Determining which peptides bind to a specific MHC classâ… molecule is not only helpful to understand the mechanism of immunity, but also to develop effective anti-tumor epitope vaccines. In order to further study the specificity of MHC classâ… molecule binding antigen peptide, the support vector regression (SVR) and four amino acid encoding schemes were used to build four models of predicting binding affinities between peptides and MHC classâ… molecules. Comparison among performances of the four models indicated that the model based on physicochemical properties of amino acids is more satisfying. Furthermore, the specificities of MHC classâ… molecule binding antigen peptide were obtained through analysis based on the contribution of the amino acids to peptide-MHC classâ… molecule binding affinities in the predictive model.3. In the exogenous antigen processing and presentation pathway, peptide binding MHC classâ…¡molecule is an important prerequisite for activating helper-T-cell mediated immune response. Accurate prediction of peptide that bind a specific MHC classâ…¡molecule is not only helpful for understanding the immune mechanism but also is useful for developing of epitope vaccine and immunotherapy of autoimmune disease, e.g. rheumatoid arthritis (RA) and Insulin-dependent diabetes mellitus (IDDM). In this paper, a method combine an iterative self-consistent (ISC) strategy with support vector regression (SVR) and four schemes of amino acid encoding was used to build models to predict binding affinities between peptides and MHC classâ…¡molecules. The predictive performance of the method is validated on data sets of 17 MHC classâ…¡alleles covering 14 human HLA DR alleles and 3 mouse H2 IA alleles. Compared to other models with the same data set, the predictive performance of our model is more satisfying. Furthermore, the specificities of MHC classâ…¡molecule binding peptide were obtained through analysis based on the contribution of the amino acids to peptide-MHC classâ…¡molecule binding affinities in the predictive model. This study is helpful to further reveal mechanism of generation of Th epitope.