Bioinformatics Analysis Of H7N9Avian Influenza Virus

To study the biological properties of H7N9subtype avian influenza virus and reveal the variation rule of avian influenza virus, we use firstly improved BP neural network and genetic algorithm to research the classification of HA and NA proteins. We propose a hyperbolic tangent function which has four adjustable parameters to improve BP neural network. Compared to traditional BP neural network, the mean square error of improved BP neural network is reduced by3%. In addition, the combination of improved BP neural network and genetic algorithm optimizes the network weights and reduces the mean square error to8%. It greatly improves the accuracy of protein classification. Secondly, we use PredictProtein software to analyze the HA protein secondary structure of H7N9avian influenza virus KC853766in Hangzhou province. Random coil structure is most likely to combine with antibody. We use DNAStar and DNAMAN softwares to analyze B cell epitopes of the HA protein and find23potential epitopes.Finally, we use clustalx and mega5softwares to make comparison and genetic evolution analysis for the nucleotide sequence of five subtypes of avian influenza which is H5N1, H9N2, H7N7, H3N2and H7N9in HA gene. Afterwards, we use Bayesian methods to predict positive selection sites of the HA protein. From the phylogenetic tree results, we find that H7N9subtype avian influenza virus may be obtained on the basis of H7N7evolution. There have31positive selection sites in HA protein. The sites which are most likely to mutate is754-756. Because211-213sites are the intersection of B cell epitopes and positive selection sites, so we infer that it may be the combat core sites of H7N9avian influenza virus KC853766in Hangzhou. This study provides a theoretical basis to further determine the B cell epitope.of HA protein and design vaccine of H7N9.