More on the 'cladistics' of sequences

[pierre deleporte]

A 13:37 07/06/2004 -0400, Herb wrote:

>It does bring up the point touch on by Zdenek Skala: If taxa are not in the
>outgroup are they then in the ingroup, or is there a limbo out there? For
>example if I find that taxa A,B,&C used as outgroups each changes the ingroup
>topology but find that C,D,&E produce the same ingroup topology, does that 
>mean
>that A,B&C are really part of the ingroup?

I fear you simply can't get the answer by counting this way. It's a 
topological question: maybe A, or B, or C+D+E is the true outgroup to the 
remaining of the taxa. The mere fact that (C,D,E) are three terminal taxa 
rooting in the same place changes nothing at all to the conclusion that 
"there is some error somewhere", but we still don't know where.
The recommendable solution is to enlarge the phylogenetic scope of the 
analysis, putting all these taxa together in the analysis as the putative 
ingroup, and picking putative outgroups outside (had I said "farther away", 
Zdenek would certainly have suggested: "phenetic criterion"!).

>  Whether building a tree this way is
>phenetic or not is an idea that I will have to think about.

I think this "counting" argument has something to do with, for instance, 
the majority rule consensus. When you have several optimal topologies for a 
data set, some researchers suggest to count the number of times one clade 
is supported and use this criterion in order to try and reduce the range of 
acceptable topologies. I think this is no argument, because each and every 
topology is as plausible as other ones. The same way, the trio "C,D,E" 
provides just a different rooting compared to that implied by "A" or "B", 
and each rooting is both as plausible and as puzzling than the two other ones.

Note that you can, and even should, put all five putative outgroups 
A,B,C,D,E together in the analysis. Outgroups can displace the relative 
position of taxa in the ingroup, but there is little argument against 
putting more relevant data in the analysis, and no argument in favor of one 
against another putative outgroup. Everything being equal otherwise, i.e. 
that you have no a priroi strong argument for discarding one or another of 
these putative outgroups as likely misleading (very doubtful alignment for 
instance).

Pierre


Pierre Deleporte
CNRS UMR 6552 - Station Biologique de Paimpont
F-35380 Paimpont   FRANCE
Téléphone : 02 99 61 81 66
Télécopie : 02 99 61 81 88