Identification of potential critical virulent sites based on hemagglutinin of influenza A virus in past pandemic strains

• Main Authors: Yin, Rui, Fransiskus, Xaverius Ivan, Zheng, Jie, Zhou, Xinrui, Chow, Vincent T. K., Kwoh, Chee Keong
• Other Authors: School of Computer Science and Engineering
• Format: Conference or Workshop Item
• Language: English
• Published: 2019
• Subjects: Engineering::Computer science and engineering; Influenza A Virus; Pandemic
• Online Access: https://hdl.handle.net/10356/102511; http://hdl.handle.net/10220/49803
• Institution: Nanyang Technological University

The influenza pandemics have caused millions of deaths and enormous economic loss. Current circulating influenza viruses in human, avian, swine and other animals are potential to evolve into novel strains that may cause another pandemic in the future. Hence, recognizing the determinants of pandemic strains helps to raise the alarm of future pandemics. With increasingly huge biological data, computational modeling is a good technique for analyzing data, providing novel insight into significant patterns and rules. Here we define a binary classification problem of categorizing influenza strains into pandemic and non-pandemic classes based on amino acid sequences. Three rule-based algorithms are applied, namely OneR, JRip and PART, to extract rules, composed of potential critical virulent sites. The results present good performance in term of accuracy, specificity, sensitivity and F-measure (more than 0.9 on average for each). Fourteen out of the sixteen potential critical virulent sites detected in our experiments are overlapped with receptor binding sites or antigenic sites. In addition, some variations occurred in these sites are known to affect the pathogenicity of influenza viruses or to cause more severe symptom in the infected patients. The pandemic potential of uncovered sites in our study needs to be further experimentally validated.