Biology Reference
In-Depth Information
is a much more complex variable than, for example, a real valued random
variable. It is composed of the discrete structure of the topology and the
continuous values of the branch lengths. Once a tree (or a set of trees) is
constructed, we want to make confidence statements about the branch-
ing patterns and the branch lengths, and moreover be able to compare
different trees or build confidence sets of trees. An elaborate treatment
of these topics is beyond the scope of this chapter. Instead, we conclude
by giving some relevant references for further reading.
Probably the best known tool to measure the robustness of a con-
structed tree with respect to small changes in the data is bootstrapping.
It is often used to obtain confidence values for the edges on a tree and
can be applied in conjunction with almost all tree building methods.
See Holmes
29
for a survey. However, statistical interpretation of the boot-
strapping supports is unclear. Alternatively, in the context of likelihood-
based methods, likelihood ratio tests
30,31
and Bayesian branch supports
32
are used.
In Sec. 3, several heuristics were described to search a tree space.
During such a search, many topologies are visited and scored. Two scores
that are different in absolute numbers need not be different in statistical
terms. Thus, another important and nontrivial issue besides branch test-
ing is the statistical comparison of different competing trees (e.g. the best
scoring tree vs. the second-best scoring tree, or trees built on different
genes). Goldman
et al
.
33
have reviewed statistical tests for the compari-
son of topologies and present their correct use.
References
1. Semple C, Steel M. (2003)
Phylogenetics
. New York, NY: Oxford University
Press.
2. Swofford DL. (2003)
Phylogenetic Analysis Using Parsimony
(*
and Other Methods
),
Version 4
. Sunderland, MA: Sinauer Associate.
3. Thompson JD, Higgins DG, Gibson TJ. (1994) CLUSTAL W: improving the
sensitivity of progressive multiple sequence alignment through sequence weight-
ing, position-specific gap penalties and weight matrix choice.
Nucleic Acids Res
22
: 4673-80.
4. Katoh K, Kuma K, Toh H, Miyata T. (2005) MAFFT version 5: improvement
in accuracy of multiple sequence alignment.
Nucleic Acids Res
33
(2): 511-8.
