Classification of human vs. non-human, and subtyping of human influenza viral strains using Profile Hidden Markov Models

Abstract

Influenza is one of the most important emerging and reemerging infectious diseases, causing high morbidity and mortality in communities (epidemic) and worldwide (pandemic). Here, Classification of human vs. non-human influenza, and subtyping of human influenza viral strains virus is done based on Profile Hidden Markov Models. The classical ways of determining influenza viral subtypes depend mainly on antigenic assays, which is time-consuming and not fully accurate. The introduced technique is much cheaper and faster, yet usually can still yield high accuracy. Multiple sequence alignments were done for all human HA subtypes (H1, H2, H3 and H5), and NA subtypes (N1 and N2), followed by profile-HMMs models generation, calibration and evaluation using the HMMER suite for each group. Subtyping accuracy of all HA and NA models achieved 100%, while host classification (human vs. non-human) has accuracies varied between (55.5% and 97.5%) according to HA subtype.