Re: Does DNA Sequence Similarity Trump Shared Chararacters?

On Sun, 17 Aug 2008 15:37:56 -0700, John Harshman
<jharshman.diespamdie@xxxxxxxxxxx> wrote:

r norman wrote:
On Sun, 17 Aug 2008 07:45:43 -0700, Jesse Ukaine <nongo10_1@xxxxxxxxx>
wrote:

On Sun, 17 Aug 2008 06:07:33 -0700 (PDT), Ron O <rokimoto@xxxxxxx>
wrote:

On Aug 16, 9:36 pm, Jesse Ukaine <nongo1...@xxxxxxxxx> wrote:
On Sat, 16 Aug 2008 17:51:22 -0700 (PDT), Ron O <rokim...@xxxxxxx>
wrote:

On Aug 16, 6:39 pm, Jesse Ukaine <nongo1...@xxxxxxxxx> wrote:
On Sat, 16 Aug 2008 15:39:10 -0700 (PDT), Ron O <rokim...@xxxxxxx>
wrote:
<snip>

I've looked at this article and it doesn't look like a very good
analysis of the data. The thing about the cladistic approach is that
you need a lot of traits to compare because a certain subgroup of
these guys claimed that they didn't have to consider evolutionary
relationships to make their analysis work. All they required was
computers to analyze the data and as many characters as they could
measure. Before computers this type of analysis was pretty much
impossible. You had to pick your characters carefully because you
could only analyze a limited number. The old school had to consider
evolutionary relationships and try to pick the most informative
character set that they could. For example they wouldn't worry about
the body shape or fins and tails in dolphins to try to figure out
their phylogenetic relationship with other mammals they would pick
features that they could compare with their terrestrial relatives.
The cladists would waste their time measuring the fins and streamlined
body shape and hope that if they measured as many traits as they could
that it would all even out and they would get some type of accurate
answer. Preliminary results might give them stupid groupings like
place dolphins with sharks or something, but as all the possible
traits got analyzed they would tend to group dolphins with other
mammals, and if they typed enough they might even be able to tell what
the closest living relatives were.
You can't just take 40 traits that fit with your notion. You have to
take all the characters and let them tell you where humans fall.
There have to be hundreds or thousands of characters that have been
compared, what can be infered from all of them?
The DNA evidence tells me that if the cladist methodology is good for
anything that it will confirm the DNA evidence with enough characters
tested.
I suppose cladistics is no different than traditional systematics in
this regard.
In traditional systematics they would spend most of their time arguing
whether a trait was useful for inferring phylogenetic relationships.
They needed the most informative set of characters to make the most
accurate inference. The cladists that I am thinking about used the
shotgun approach and figured numbers would win out. This guy with his
40 characters seems to be taking a more traditional approach so he has
to be able to argue that his characters are informative. Just look at
a marsupial mole and a eutherian mole from above. Just looking at
them you might be fooled into thinking that they are closely related,
but as soon as you looked into their reproductive system you would
know that around 180 million years of evolution separate the two. You
obviously can't just take a limited number of characters and expect to
get an accurate answer unless you prescreen and pick the most
informative ones.

The cladists with the shotgun approach should love DNA analysis. It
gives them literally thousands of characters to analyze. Between
chimps and humans we have millions. DNA analysis is subject to many
of the same problems. Small data sets can give you bogus results.
Some sequences are known to be hypervariable and aren't good for
certain comparisons, we have to deal with things like horizontal
transfer, etc., but if you swamp the analysis with numbers you should
get a fairly accurate interpretation. A scientist can literally get a
group of blood sample or DNA, and without knowing what the animals
look like, correctly group them by genetic relationships. If he has a
database to compare the sequences to he can even tell what they are
most likely to be. That has about as little of the evolutionary bias
that the cladists were trying to avoid as any analysis that I know of.
OK. Thanks. I guess this clarifies why you guys kept saying things
like "Listen up, idjit, DNA sequences *are* characters!"

One of the more important factors going on here is the sociological
process occurring in the changes in the nature of biology in the past
few decades. Evolution had for ages been the domain of morphologists
who were trained in certain areas now called "classical biology" to be
polite and "old fashioned stamp collecting" to be rude. The DNA
business came out of molecular biology and extremely arrogant and
brash young kids who proclaimed that they knew everything. James
Watson's attitude made evident in his "Double Helix" autobiographical
account is part of that. There was an enormous amount of antagonism
between the groups as the political structure of older Zoology,
Botany, and Bacteriology departments became reorganized into newer
Biological Science groups. The older group lost a tremendous amount
of power (read lab space and financial support). Modern (that is, in
the last ten to twenty years) systematic biologists are, of course,
now well versed in molecular techniques and are happy to deal with DNA
sequences. The fact is that evolutionary relations must be based on
what is inherited and what is inherited is DNA sequence, not
morphological structure. There still remain clusters of individuals,
though, who stick to the old ways and try to retain some level of
prominence to purely morphological characters. There is validity to
that approach, also. The fact is that fossils do not produce DNA
sequences, at least no really old stuff, and morphology of hard parts,
things like skulls and teeth of vertebrates for example, are pretty
much all we have in many cases.

Morphological analysis is also very useful in giving us something to do
with all those trees we get from molecular characters. If you want to
use those trees to learn something about morphological evolution, you
have to study the morphologies of the various species, score their
characters in some way, and hang them onto the trees. Comparative
anatomy, dontcha know.

Actually, as an organismal biology I am all in favor of morphological
analysis. It is just that I also know which side of the bread the
butter is on and whose hand feeds me. Comparative anatomy (and
physiology!) rules but just don't say it too loudly.

.

Loading