Human influenza: a virus classification using a probabilistic neural network

Abstract

Purpose

The traditional method to distinguish the serotype of influenza A virus was based on the antigen reaction of HA and NA with their antibodies. The antibody of specific subtype virus was difficult to get and the reaction was not easy to be done. To be a complementation to the traditional classification methods based on the serological reaction, aims to present a novel method for viral classification.

Design/methodology/approach

The similarity values of all subtype HA genes in vector space were considered and classified using a probabilistic neural network (PNN). The PNN model was trained by the 132 viral sequences in the training set and the classification quality was examined using 28 viral sequences in the testing set.

Findings

A novel technique for the serotype classification of influenza A virus isolated from human was proposed in the paper. The system achieved 100 per cent accuracy with all serotypes of human influenza A virus.

Research limitations/implications

The time for the large‐scale calculations of the average similarity based on multisequence alignment is the main limitation.

Practical implications

This is a supplementation to the traditional virology research.

Originality/value

The novel classification method based on the similarity of viral nucleotide sequences and the PNN model would be useful for the epidemic supervision and prevention.

Citation

Kou, Z., Zhou, Y. and Qiang, X. (2008), "Human influenza: a virus classification using a probabilistic neural network", Kybernetes, Vol. 37 No. 9/10, pp. 1425-1430. https://doi.org/10.1108/03684920810907733