Prediction Of CTL Epitopes Based On Modified Artificial Neural Network

The immune system of the human beings is activated through the immune response to have the potential of immune defense,immune homeostasis and immune surveillance so that it can adjust to the external environment and keep the equilibrium and steady of the internal. The process of the immune responses are executed by the cell immunity and humoral immunity induced by the T cell and B cell respectively.The paper focuses on the cell immunity.The antigen peptide protein of the tumor can be degraded to many polypeptides with different length by the proteasome or non-proteasome of the antigen processing cell(APC)or target protein.Then some of the polypeptides are selectively loaded by transporter associated with antigen processing to endoplasmic reticulum.With the assistance of the chaperons such as calnexin,calreticulin and tapasin,the molecular heavy chain andβ2 micro globulin of the major histocompatibility complex(MHC)are assembled together to form the steady complex which is further transported to the surface of the APC through the Golgi body.The binding of the MHC classⅠto antigen peptide is a pivotal step in the immune response of the specific cell.It proves that most CTL epitopes can bind to the MHC classⅠmolecular with high or medium affinity.In conclusion,the MHC classⅠpathway of the antigen peptide processing and presentation is a multi-steps and complicated process.It has three essential steps:the cleavage production by the proteasome,transported to the ER by the TAP molecular and binding to the MHC classⅠmolecular.With the development of the bioinformatics and computer technology,many theoretical methods are used to make the prediction of the CTL epitopes,such as statistical classifiers,neural classifiers,heuristics-based,structural classifiers and molecule modeling method.And they all have great potential to develop and improve.The paper quantitatively studied the binding affinity relationship between peptide and MHC classⅠmolecular with the modified artificial neural networks method.The model was established to predict epitopes of cytotoxic T lymphocyte.The affinity prediction with this model to the 805 peptides samples of HLA-A*0201 molecular was carried on.The accuracy of prediction achieves 73.8%.To peptides came from MAGE-2 protein,the binding affinity prediction was also carried on,and the predicting result is satisfying.