Rooting a phylogenetic tree with nonreversible substitution models
1 Mathematics Department, University of California, 970 Evans Hall, Berkeley, CA 94720, USA
2 Statistics Department, University of California, 367 Evans Hall, Berkeley, CA 94720, USA
3 Division of Genetics and Bioinformatics, The Walter & Eliza Hall Institute of Medical Research, 1G Royal Parade, Parkville, Vic 3052, Australia
BMC Evolutionary Biology 2005, 5:2 doi:10.1186/1471-2148-5-2Published: 4 January 2005
We compared two methods of rooting a phylogenetic tree: the stationary and the nonstationary substitution processes. These methods do not require an outgroup.
Given a multiple alignment and an unrooted tree, the maximum likelihood estimates of branch lengths and substitution parameters for each associated rooted tree are found; rooted trees are compared using their likelihood values. Site variation in substitution rates is handled by assigning sites into several classes before the analysis.
In three test datasets where the trees are small and the roots are assumed known, the nonstationary process gets the correct estimate significantly more often, and fits data much better, than the stationary process. Both processes give biologically plausible root placements in a set of nine primate mitochondrial DNA sequences.
The nonstationary process is simple to use and is much better than the stationary process at inferring the root. It could be useful for situations where an outgroup is unavailable.