[Taxacom] Why character-tracking doesn't happen?

Bob Mesibov mesibov at southcom.com.au
Fri Sep 12 23:16:32 CDT 2008


Pierre Deleporte wrote:

"- suggestion for saving computing time in your approach: if you
consider 
that homoplay is impossible for some characters, you better introduce a 
constraint in your analysis so that any topology  implying homoplasy
for 
these characters is discarded from the beginning - they can't pop up as 
a result, your problem is limited to choosing among the remaining set
of 
topologies, possibly very limited in some cases (easy job if it boils 
down to zero except a big unresolved radiation, an efficient protection 
against any possible homoplasy - another being clique analysis: retain 
only the subset of completely congruent characters and publish a nice 
CI=1...)"

But there are already serious constraints in any phylogenetic analysis.
They are put there by optimising, and most of them aim to minimise
homoplasy. I don't see the value of saying that for some characters, I
will minimise to zero. Mind you, that already happens with nucleotide
models. Penalties for some changes are so great that they are
effectively impossible.

Another way to look at this matter is not to ask 'How do we test
trees?', but simply to accept those branching arrangements which appear
again and again in analysis after analysis and are therefore more or
less noise-free (i.e., free of noise from sampling or analytical error;
there will always be genuine 'evolutionary noise'), and then see what
character-change patterns are inherent in those topologies.

For example, we have been assured for many years now that molecular
phylogenies based on genes not subject to intense selection pressure,
such as the ribosomal RNA genes, should be robust and credible. Suppose
that to be true. If we then map onto a very solid consensus molecular
tree (consensus of many independent genes) some morphological characters
of interest, do we see robust and credible histories of those
morphological characters, including reversals, likely convergences and
likely parallelisms? (Post-analysis mapping of this kind is routinely
done with biogeographical data, of course.)

Well, this is not a popular approach. More often a morphological data
set is analysed independently and its result 'consensus'ed' with the
molecular tree, or the published tree is based on a combined morphology
+molecules (apples+oranges) analysis. The aim in both cases is to
produce the best possible tree using 'all' the available evidence. Why?

[Logically, we should not look to such a tree to track the morphological
characters already used in the analysis, but this raises the nasty
question of how to discover characters of interest which evolve
independently of those previously analysed.]
-- 
Dr Robert Mesibov
Honorary Research Associate
Queen Victoria Museum and Art Gallery and
School of Zoology, University of Tasmania
Home contact: PO Box 101, Penguin, Tasmania, Australia 7316
(03) 64371195; 61 3 64371195
http://www.qvmag.tas.gov.au/mesibov.html





More information about the Taxacom mailing list