Linguistic analysis of the nucleoprotein gene of influenza A virus

Applies a linguistic analysis method (N-grams) to classify nucleotide and amino acid sequences of the nucleoprotein (NP) gene of the influenza A virus isolated from three hosts and several geographic regions. We considered letter frequency (1-grams), letter-pairs' frequency (2-grams) and triplets' frequency (3-grams). Nearest-neighbor classifiers and decision-tree classifiers based on 1-, 2- and 3-grams were constructed for NP nucleotide and amino acid strains, and their classification efficiencies were compared with the groupings obtained using phylogenetic analysis. Our results show that disregarding positional information for NP can provide almost the same high level of classification accuracy as alternative, more complex classification techniques that use positional information.