Learning evolutionary and virulence patterns of influenza viruses

• Main Author: Tan, Tosy Ying Jie
• Other Authors: Kwoh Chee Keong
• Format: Final Year Project
• Language: English
• Published: 2019
• Subjects: DRNTU::Engineering::Computer science and engineering
• Online Access: http://hdl.handle.net/10356/76446
• Institution: Nanyang Technological University

Without any warnings, influenzas can strike and take away the precious lives of both humans as well as livestock. They are deadly, uninvited and severe. It still remains as a blur amongst experts on how such viruses can actually result into an epidemic or pandemic. The project that I embarked on is a continuation of a current Final Year Project (FYP). This project aims to study on the virulence patterns of influenza viruses, hoping to bring us one step closer to being ahead in the race of evolution of viruses where we will be able to predict the possibility of a pandemic before it actually occurs. Focusing on Type A and B influenzas, the median lethal dosage value was explored to see how it affects the virulence level of the viruses. A series of data processing steps has to be done before classification can take place. Three different classification methods, namely JRip, OneR and PART, were used in this project. The influenza viruses were first classified according to the various ribonucleic acid (RNA) segments. The classification metrics that returned the best results out of the three cases tested was further explored where different types of categorisation of the dataset (eg. by host strains and subtypes) were considered. Boosting techniques were also applied to further improve the classification results. Although the classification results were not as ideal, we managed to conclude that the median lethal dosage has an influence in determining the virulence level of an influenza virus. It was also proven that the HA segments contains crucial information on the virulence of influenza viruses as well as shown that categorisation by host strains seems to produce better classification results. Hence, further works recommended can be to (1) narrow down the scope of focus (eg. solely by subtype or RNA segment) first to have a better and more complete understanding on the virulence patterns or (2) further tuning parameters for boosting to improve the classification performances.