Prediction of interspecies transmission for avian influenza A virus based on a back-propagation neural network

Avian influenza A viruses could cross the barrier and infect human beings. We need to predict which subtypes and strains of avian influenza virus will become capable of interspecies transmission, which is very important for public health. Recently, two categories of molecular patterns, which were associated with the phenotype of interspecies transmission, were found to exist among avian influenza A viruses with different subtypes. In this study, we used a method based on wavelet packet decomposition to transform the viral sequences and used the energy features of the viral genome as input to train a back-propagation (BP) neural network. The average output values for test samples were 0.0125 and 0.0092. The good results of the experimentation were used to discriminate the two categories of molecular patterns. A novel method was constructed to predict the transmissibility of avian influenza A viruses, which is very useful for public health. It was also anticipated that the current classification method would be a useful tool based on the development of large-scale genome sequencing of the influenza A virus.