The Research Of T-cell Epitope Prediction Using The Method Of PCA On Clustering Result

During the process of immune response. T cell cannot identify the complete naturalantigen directly but need the help of T cell Receptor on the surface of T cell to identify apeptide with given function in the antigen protein. This peptide is referred to as a T-cellEpitope or Antigenic Determinant. Major Histocompability Complex(MHC) can be deviedinto MHC class I and MHC class II according to the differences of their function. Theyparticipating into endogenous and exogenous antigen presenting process respectively. Topredict the T-cell epitope interacting with MHC molecule not only have contributes to anunderstanding of the principle of the immune response of the diseases such as autoimmunediseases, anaphylactic reaction, infectious disease and tumor, but also have very importantsignificance for computer-aided design of artificial vaccines and immune intervention. Atpresent, the positioning method of the T-cell epitope need to consume a great deal ofmanpower and material resources and have high requirements of the equipment.As the increasing of the known epitope data, the computer, an effective means ofauxiliary experiment, is applied in biological experiments by researchers. The waitingepitopes obtained from the computer predictor can be verified by the subsequent biologyexperiments. Combine the computer predictors and biology experiments felicitously canensure the accuracy of result at the same time to save the cost and improve the efficiency toproduction needs of modern society.So far, more effective methods to predict the epitope interact with MHC molecule areBinding Motif, Quantitative Binding Matrix and Machine Learning. In this article, we comeup with a method of PCA on clustering result to reduce the physicochemical properties ofamino acid which participated in the process of prediction. Thus we can select more relevantphysicochemical properties with given MHC molecule to improve the performance of thepredictor.The experimental results show that the method proposed in this paper can improve theperformance of the predictor effectively.