Tracking the geographical spread of avian influenza (H5N1) with multiple phylogenetic trees
Version of Record online: 13 NOV 2009
© The Willi Hennig Society 2009
Volume 26, Issue 1, pages 1–13, February 2010
How to Cite
Hovmöller, R., Alexandrov, B., Hardman, J. and Janies, D. (2010), Tracking the geographical spread of avian influenza (H5N1) with multiple phylogenetic trees. Cladistics, 26: 1–13. doi: 10.1111/j.1096-0031.2009.00297.x
- Issue online: 5 JAN 2010
- Version of Record online: 13 NOV 2009
- Accepted 6 October 2009
Avian influenza (H5N1) has been of great social and economic importance since it first infected humans in Hong Kong in 1997. A highly pathogenic strain has spread from China and has killed humans in east Asia, west Africa, south Asia, and the Middle East. Recently, several molecular phylogenetic studies have focused on the relationships of various clades of H5N1 and their spread over time, space, and various hosts. These studies examining the geographical spread of H5N1 have based their conclusions on a single tree. This tree often results from the analysis of the genomic segment coding for the haemagglutinin (HA) or neuraminidase (NA) proteins and a limited sample of viral isolates. Here we present the first study using multiple candidate trees to estimate geographical transmission routes of H5N1. In addition, we use all high-quality HA and NA sequences available to the public as of June 2008. We estimated geographical transmission routes of H5N1 by optimizing multistate characters with states representing different geographical regions over a pool of presumed minimum-length trees. We also developed means to visualize our results in Keyhole Markup Language (KML) for virtual globes. We provide these methods as a web application entitled “Routemap” (http://routemap.osu.edu). The resulting visualizations are akin to airline route maps but they depict the routes of spread of viral lineages. We compare our results with the results of previous studies. We focus on the sensitivity of results to sampling of tree space, character coding schemes, optimization methods, and taxon sampling. In conclusion, we find that using one tree and a single character optimization method will ignore many of the transmission routes indicated by genetic sequence and geographical data.
© The Willi Hennig Society 2009.