More on the 'cladistics' of sequences

[pierre deleporte]

A 18:34 06/06/2004 -1100, John Grehan wrote :
>I did not say character states are determined in the absence of knowledge
>of phylogenetic relationships! To the contrary, I am always referring to
>evaluation of each character with respect to an outgroup before the analsis.

And you now know that this is exactly what programs do when they perform 
cladistic analysis of molecular data like for any other data. I know that 
you know that. Via the outgroup criterion, each and every molecular 
character state (i.e. a given base at a given site in aligned sequences) 
has an a priori polarization in putative plesio-apomorphy (for each and 
every character = site).

And you also know that there is nothing like "rooting after the analysis", 
or "rooting during the analysis", because you know that one can perform the 
analysis the following way and get exactly the same result:

- first, root optimally all possible topologies according to the data and 
outgroup criterion
- second, pick out the optimal topology (maximizing homology).

You'll get exactly the same result than with applying the following procedure:

- first, pick out the optimal unrooted topology (maximizing homology)
- second, root it optimally using the outgroup criterion.

And you also know that the program begins with discarding cladistically 
non-informative characters. This is the first thing it does. Thus, only 
cladistically putatively informative characters remain in the analysis, 
i.e. characters with putative plesiomorphic state and apomorphic ones. The 
data matix is effectively reduced to these characters, and they are not 
treated phenetically (i.e. grouping otaxa on the basis of overall 
similarity). And only apomorphies play a role for polarizing the tree : 
this is what optimal outgroup rooting does.

Hence, all outgroup-polarized molecular characters are "cladistic" in your 
(very peculiar) acception of the term, i.e. they are individually, 
putatively polarized a priori, via the outgroup criterion, and the analysis 
is performed the classic cladistic way for molecules just like for 
morphology. Same criteria, same procedures.

Of course Hennig did not use a computer, and did not apply systematically 
(via an algorithm) his "auxiliary principle" of preferring interpretation 
in terms of homology rather than homoplasy (what is implemented through the 
"congruence criterion" for choosing the optimal unrooted topology).

But this doesn't make modern cladistics "non-hennigian" in this respect, 
and molecularists "know" their ingroups and outgroups just like the 
morphologists do, no more no less, and they face the same problems in this 
respect (possible problem of multiple rooting for multiple outgroups, whose 
solution consist in enlarging the phylogenetic scope of the analysis and 
using more data), and I still cannot understand why you persist in taxing 
molecular cladistic phylogeny of being non-cladistic.

>One can document each character for the outgroup and ingroup. By this
>documentation it is possible for each character to be independently
>verified or refuted by another individual

This you can do, exactly this, with molecular data as treated by modern 
programs. Just try it, as I suggested you repeatedly. But apparently you 
don't try... Why don't you try and verify by yourself that this is all the 
same approach? Same logic giving same result?
I admit that the fact that everybody tells you the same thing will not 
change your mind the slightest way, for it's quite imaginable that the 
whole community of specialists of morphological and molecular cladistic 
analysis on earth is wrong and you are right. Science is not a democraty. 
But why don't you try and verify? Because it's also imaginable that you are 
wrong, and this you can check by yourself:
- take some molecular data
- root them a priori character by character via the outgroup criterion
- throw away putatively cladistically non-informative characters (obvious 
autapomorphies and unchanging characters)
- find the optimal tree your own cladistic way
- try with the program PAUP using the same assumptions (costs of changes...)
- check if you get the same result.

>No, but if one cannot polarize the characters and determine which are
>potential synapomorphies before the analysis then the implication is that
>such individuals do not know their group very well.

But this is exactly what the program does, and of course the program itself 
does not even knows whether the data you feed it are morphological or 
molecular. How could molecular cladistics be different from morphological 
cladistics ? Any molecular analysist can tell a priori what are the 
potential apomorphies: just like for morphology, they are the character 
states not present in the outgroup(s). You said "potential": this is the 
rigth term (or "putative"), because some putative plesiomorphies in the 
outgroup(s) may finally appear as autapomoprphies of this outgroup. But 
this doesn't change anything to the optimal rooting of the optimal topology 
for a given data set.
If you're not convinced, just try it, once again, and check.

>I would start with some critical evaluation before the analysis to restrict
>the data set to potential synapomorphies.

All characters with more than one state have potential plesio-apomorphy 
polarity. The outgroup provides the support for polarizing. And the program 
makes no use of cladistically potentially non-informative characters. So, 
where is the problem?

>Then the evaluation after the
>analysis can take place with respect to one's initial determination.

This is what contemporaneous computer-assisted cladistic analysis, followed 
by secondary checking of optimal scenarios for characters, is all about. 
Once more, where is there any problem with molecular data?

Best,
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