The hemagglutinin mutation E391K of pandemic 2009 influenza revisited.

Phylogenetic analyses based on small to moderately sized sets of sequential data lead to overestimating mutation rates in influenza hemagglutinin (HA) by at least an order of magnitude. Two major underlying reasons are: the incomplete lineage sorting, and a possible absence in the analyzed sequences set some of key missing ancestors. Additionally, during neighbor joining tree reconstruction each mutation is considered equally important, regardless of its nature. Here we have implemented a heuristic method optimizing site dependent factors weighting differently 1st, 2nd, and 3rd codon position mutations, allowing to extricate incorrectly attributed sub-clades. The least squares regression analysis of distribution of frequencies for all mutations observed on a partially disentangled tree for a large set of unique 3243 HA sequences, along all nucleotide positions, was performed for all mutations as well as for non-equivalent amino acid mutations - in both cases demonstrating almost flat gradients, with a very slight downward slope towards the 3'-end positions. The mean mutation rates per sequence per year were 3.83×10(-4) for the all mutations, and 9.64×10(-5) for the non-equivalent ones.