More on the 'cladistics' of sequences

[Richard Jensen]

John Grehan wrote:

> At 01:24 PM 6/11/04 -0500, Richard Jensen wrote:
>
> >That's my point - the character could be used in either a phenetic or a
> >cladistic analysis.  The same data matrix can be analysed by a cladistic
> >algorithm or a phenetic algorithm.
>
> Well I can agree with this although I would say that the latter would not
> be cladistics if the characters were not already limited to predicted
> synapomorphies.

By latter, I assume you mean my reference to a phenetic algorithm.  These
generally are not intended to provide cladistic interpretations in the first
place (except for some methods, e.g., neighbor-joining) , so I'm not sure what
the intent of your comment is.  As I noted several months ago, and others have
more recently noted, there are corcumstances under which a phenetic analysis
may yield results hat are a good reflection of phylogeny.  However, a
fundamental difference between phenetic (here is use the term exclusively for
algorithms based on overall similarity) and cladistic (here I use the erm
exclusively for methods such as that developed by Hennig) analyses is that the
former does not treat characters as discrete entities with respect to
determining relationships while the latter does.  One can map discrete
characters on a tree derived from a phenetic analysis, but it is likely to
illustrate more homoplasy than a cladistic tree derived for the same data
matrix (as suggested by Sokal and others; real counter examples do exist).  You
may create a data matrix limited "to predicted synapomorphies;" that matrix may
be analyzed by any tree building protocol; the results will almost certainly be
different as a function of the different ways the relationsghips among the taxa
are evaluated and the tree is constructed.

>
>
> >The results will most likely be
> >different, as will the interpretation, but that's a function of how the
> >characters are analyzed, not how they are are coded.
>
> I agree that the results may be different depending on what algorithm is
> used, but I also assert that the results can be different depending on what
> kind of characters are used. Using characters that are an undetermined mix
> of primitive and derived characters is, to me, just a phenetic analysis
> whether or not a cladistic algorithm is used on the data.

This is a point Pierre and I have been trying to make - if the matrix consists
of a mix of "primitive and derived characters," the cladistic algorithm will
sort them out for you and you'll have a cladistic result.  A cladistic
algorithm will not yield a phenetic result.

> So the lack of
> correspondence between the huaman-orangutan morphology and DNA sequences
> could be due (with respect to the DNA sequences) that:
>
> 1. The DNA sequence data only represents an overall similarity of DNA
> sequences and is therefore not a necessary match for phylogeny;
> 2. And DNA sequences may not necessarily match the genetics connected with
> speciation events that is represented in the morphological synapomorphies.
>
> John Grehan
>

The morphological data could also be reflecting long-branch attraction because
there are parallelisms between the orang-human lineages and these (apparent)
parallelisms may not be reflected in the DNA data (and the reverse could happen
as well).

Dick

--
Richard J. Jensen              | tel: 574-284-4674
Department of Biology      | fax: 574-284-4716
Saint Mary's College         | e-mail: rjensen at saintmarys.edu
Notre Dame, IN 46556    | http://www.saintmarys.edu/~rjensen