The evasion of influenza virus from host immune surveillance is mainly mediated through its
surface protein hemagglutinin (HA), the main component of
influenza vaccine. Thus, identification of influenza virus antigenic
epitopes on HA can not only help us understand the molecular mechanisms of viral immune escape but also facilitate
vaccine strain selection. Despite previous efforts, there is a lack of systematic definition of the antigenic
epitopes for the highly pathogenic
avian influenza (HPAI) H5N1 viruses. In this study, we infer the HA antigenic
epitopes for H5N1 viruses by integrating the antigenic sites mapped from the HA of
human influenza H3N2 viruses, the sites which were reported to be associated with immune escape in H5 viruses and the mutation hotspot sites identified in the evolutionary history of HPAI H5N1 viruses. We show that these inferred antigenic
epitopes play significant roles in antigenic variation of HPAI H5N1 viruses. Based on inferred antigenic
epitopes, we further develop a computational method to effectively predict antigenic variants for HPAI H5N1 viruses (available at http://biocloud.hnu.edu.cn/predict/html/index.html). Therefore, our work has not only inferred the antigenic
epitopes for HPAI H5N1 viruses but also provided an effective computational method to assist
vaccine recommendations for protection against the deadly bird flu.