Developing T helper cells/major histocompatibility complex molecules feature encoding methods in detection of binding sites

T cells, has an important role in the formation of immune response. The human body constantly faces with from the outside and ever changing a large number of microorganisms. Antigen proteins having the harmful microorganisms cause the activation of the immune system. Long antigen proteins are divided into smaller peptide fragments by antigen presenting cells to combine with T cells. This antigenic peptides are defined Epitope. Different T cell clones to each peptide is considered to be an existing in nature of antigens on protein structure. Another important feature of the T cells must be processed to recognize and create reaction the antigen formed by some cells and offered them via surface molecules. The major histocompatibility complex is called enabling T cells to molecules on the cell surface antigen presentation. Connecting the major histocompatibility complex molecule leads to T cell activation and triggers a series of biochemical reactions. The prediction of T cell epitopes identification is important to develop a vaccine or drug on the immune system. Identification antigenic peptides in excess of one thousand varied undergoes sustained mutations is not appropriate in terms of time and cost in the laboratory. Therefore, it is more appropriate to seek solutions with computerized machine learning algorithms. Develop a new machine learning techniques to predict attributes encoding epitopes is the purpose of fhis thesis. IEDB database of human leukocyte antigen (ILA-A, ILA-B) peptide data was used for the identification T helper cells / BDUK molecule binding specificity. The data set consists of peptides nine amino acids in length. Two attribute encoding methods was developed to detect T helper cells / BDUK molecule originality. In the first method the physicochemical properties of the amino acid substitution matrix with Blosum 50 was used. In the second method the weight and the position information of amino acids and Blosum 50 substitution matrix was used. Classification tests were carried out according to the 10-fold cross-validation test technique with Weka software environment. Experimental studies in the thesis were obtained class accuracy, sensitivity, specificity and Matthews Correlation Coefficient (MKK) of performance metrics.