Goldman Group Projects
Statistical Tests of Phylogenetic Topologies
This page contains information relating to a review paper on likelihood-based
statistical tests of competing hypotheses of evolutionary relationship
Likelihood-based tests of topologies in phylogenetics.
Systematic Biology 49:652-670.
[click here to request electronic reprint]
If you or your
institution have an electronic subscription, you may be able to get
a copy of the paper via this link.
In this paper, we contend that the majority
of published applications of the Kishino-Hasegawa (1989) test (as
implemented, for example, in the PHYLIP (Felsenstein, 1995), PAUP*
(Swofford, 1998), PUZZLE (Strimmer and von Haeseler, 1996) and MOLPHY
(Adachi and Hasegawa, 1996) software packages) are using the test
for purposes for which it was not intended and for which it is inapplicable.
Further, the results of such inappropriate tests may be misleading.
The good news, however, is that alternative tests are available.
These exist in both parametric (Swofford et al., 1996) and non-parametric
(Shimodaira and Hasegawa, 1999) forms. We have prepared some
notes on using the Seq-Gen and PAUP* programs to perform parametric
tests (last updated in 2000); these notes are available from this
link. The notes should be read in conjunction with the paper
All this, with examples, is described fully in our paper.
Andrew Rambaut's SHTests software and Ziheng Yang's PAML
package implement the non-parametric test of Shimodaira and
Hasegawa (1999). Hidetoshi Shimodaira's CONSEL package is
designed to perform this (and other) tests based on the
output of various popular phylogenetics packages.
Two data sets referred to in the paper are available from the links below:
HIV-1 Subtypes A, B, D and E gag and pol Nucleotide Sequences
contains: 6-sequence multiple alignment of HIV-1 nucleotide
sequences from four subtypes. There are 2000 columns in the alignment.
The sequences are named Q23 (referred to as A1 in the paper cited above),
U455 (A2), BRU (B), NDK (D), 90CF11697 (E1), 93TH057 (E2). Please
contact Jon Anderson (email@example.com)
for further details.
Mammalian Mitochondrial Protein Amino Acid Sequences
contains: 6-sequence multiple alignment of mammalian
mitochondrial protein amino acid sequences. There are 3414 columns
in the alignment. The sequences are named Homsa (human), Phovi
(harbor seal), Bosta (cow), Orycu (rabbit), Musmu (mouse), Didvi
(opossum). Thanks to Hidetoshi Shimodaira for giving us the data
in this format. Please contact Nick Goldman (firstname.lastname@example.org)
for further details.