[Taxacom] base alignment

Vitor Miranda vmiranda at umc.br
Wed Aug 30 10:05:18 CDT 2006

Dear Pierre and all,

    I am not sure about some ideas. I guess some of the questions here 
in discussion lie on a different field than semantic comparisons between 
"phenetic" vs. "phyllogenetic" concepts.
    Primary homology (using or not embriology, ontogenic studies, 
"hypothetical ancestors", etc.) is always inferred by overall similarity 
(be nucleic bases, morphology, etc.).  Always the first (but not only) 
evidence will be the similarity. So the primary question is, when using 
molecular data like bases, how to infer similarity? That's a very 
important point, in my opinion so ignored considering the the 
computational methods. I am not so sure if most of them, but some of the 
computing aplicatives (very in use) are as a "Pandora box". Once they 
use a distance matrix as a start point, the problem is the sensitivity 
to the order of the sequences (taxa). If sequences with to much possible 
gaps or with low similarity (below 70%)  were aligned, the pairwise 
method used to several programs can return a computational problem: if 
the same sequences were offered to the aplicative in differents 
positions (to a test matrix with 10 taxa, try to invert the sequences 
positions, in one matrix using the sequences 1,2, 3 to 10 and another 
matrix inverting the same sequences in 10, 9, 8 to 1), defferents 
alignments can be produced.
    So, I think that's a great reason to try using parsimony methods to 
make alignments. If different alignments (primary homology) could be 
construct using the same data, why not choose the most parsimonious one? 
Of course if one also uses the parsimony view to create the cladogram.
    Just my two cents.
    All the best,

pierre deleporte wrote:

>A 12:28 29/08/2006 -0400, John Grehan wrote:
>>I am just giving my impression of what I have
>>read so far. For example, one paper I read stated that alignment uses
>>distance measures which to my understanding is not cladistic.
>Alignment can effectively be done on the basis of overall similarity. But 
>please note that this is not data analysis, just primary homology 
>assessment, and also please read other papers, and check for "alignment 
>corrected by hand", or other more or less sophisticated alignment methods.
>"By hand" means that molecularists are using fairly the same criteria as 
>morphologists for homology assessment: detecting optimal "connections" in a 
>molecular sequence, like for connections of a bone in a squeleton. After 
>all one could state that squeletons of different species are "aligned by 
>overall similarity" some way, with some corrections (a bone is likely 
>missing here, some are likely fused there...)...
>You certainly can see that if you really want.
>>In the sense that parsimony is the shortest tree then that is a concept
>>not restricted to cladistics.
>Finding the shortest tree and rooting on outgroups is merely the computing 
>shortcut for finding exactly the optimal cladistic tree you are obtaining 
>by hand on your morphological data (be it with compatibility or standard 
>parsimony methods).
>You can also see that if you really want (just check basic courses in 
>cladistics, or even simply the user's guide of PAUP to start with).
>> I am not saying that molecular trees cannot be correct, just
>>that they are not necessarily correct just because they use DNA
>>sequences, and that morphology is not necessarily wrong when it does not
>>corroborate molecular trees.
>OK but this is clearly not the point at stake when you are globally 
>charging all cladistic molecular analyses of being "phenetic" - a nonsense 
>in itself.
>Beside your esoteric acception of "cladistic characters" (= surviving a 
>compatibility analysis), hence of "phenetic characters" by the way (= any 
>other kind of data set), your charge that molecular phylogenies are always 
>based on overall similarity is flatly wrong.
>Ze fact iz zat cladistic ones are cladistic, you know... yes, exactly like 
>for morphological data... I'm pretty sure you can see that (if only...).
>Pierre Deleporte
>CNRS UMR 6552 - Station Biologique de Paimpont
>F-35380 Paimpont   FRANCE
>Téléphone : 02 99 61 81 63
>Télécopie : 02 99 61 81 88
>Taxacom mailing list
>Taxacom at mailman.nhm.ku.edu

Prof.Dr. Vítor Fernandes Oliveira de Miranda
Professor Assistente
Universidade de Mogi da Cruzes - UMC
Laboratório de Sistemática Vegetal
Herbarium Mogiense - HUMC (Sala 21.16 - Prédio II)
Av. Dr. Cândido Xavier de Almeida Souza, n.200
CEP 08780-911 - Moji das Cruzes - SP - Brasil
Tel.: +55 (11) 4798-7000 Ramal 7313
e-mail: vmiranda at umc.br

Taxacom mailing list
Taxacom at mailman.nhm.ku.edu

More information about the Taxacom mailing list