Prediction Of Continuous B-cell Epitopes Based On Modified BP Neural Network

With the human genetic engineering developing and genome sequence data increasing, a new interdisciplinary subject-bioinformatics is formed. As all biological data couldn't be experimental validated, it is important that experimental data analysis, sorting and effectively prediction in order to make more effectively experiment and make full use of limited experiment resources.B lymphocyte is one of important immune cells. They differentiate and mature in the bone marrow. With the Th cell help, B cells specific bind antigen in the peripheral lymphoid tissue, and then differentiate plasma cells, secrete antibody, open their immunology ability. The important symbol of B cells maturation is that IgM, IgD and BCR which composed by Iga/IgB are expressed in cell membrane. Immunoglobulin IgM and IgD recognize the antigen specifically and bind antigen, and then transmit the information through the electrical signal to the Iga/IgB chain. Iga/IgB chain deliver the signal to the cells to promote the further differentiation of B cells to achieve the immune response. The immune response mediated B cells will start only after they recognize antigen, and then B cells could express their immunity. It is shows that the antigens in the immune system play an important role. The antigen is protein fragment, which refers to material binding antibodies or immune cells receptors in the immune response, and it is the key to activate the immune response. Usually, the antigen specific binding to B cell is called B cell epitopes. It can be seen that the B cell epitope prediction is particularly important.B cell epitope divide to continuous epitopes and non-continuous epitopes, For non-continuous epitopes prediction, it need to determine the three-dimensional structure of antigen, so there are great difficulties. Presently, most international studies are continuous B cell epitopes theoretical prediction. In order to make B cell continuous epitopes prediction quickly and efficiently and improve the success rate of identification experiments, this paper applies the improved BP neural network to B cell continuous epitope theory prediction, and eventually establishes the B cell epitopes predictive model. Compared to the existing home and abroad prediction model, this model has better prediction performance (AUC=0.723). To further estimate the performance of the model, The circumsporozoite protein were predicted, and the result is satisfied.