| A neural network classification method，a batch-learning self-organizing map（BLSOM），was established using 25 tri- and tetranucleotide in the hemagglutinin gene sequences of 25 avian influenza viruses isolated from Guangdong province. Statistics and normalization of the fragment numbers were done and MATLAB function was used to simulate the human brain thinking for self-organizing learning. When the training step was 100 and above，the strains could be successfully clustered. H1，H3, H5, H7 and H9 subtype strains were mainly classified as a class, in which the HA gene cluster profiles of H3N2 and H7N9 strains had highly similar，suggesting that these isolates origin ated from the same era；different generations of H5N1，H1N1 and H9N2 strains were quite different; each one strain of two types was clustered into one group，suggesting that the two virus had recombinant variants, to provide reference for screening high-risk strains and traceability. |
