Molecules and Convergence [was: Re: Ailurus, the Red Cat]

["Jonathan R. Wagner" <jonathan.r.wagner@mail.utexas.edu>]

Jaime wrote:

> Convergence is often strange, but
> it is harder to imitate molecules than morphology, perhaps.

Let's not be too hasty, DNA does some pretty weird things that "regular
morphology" generally does not do. For one thing, DNA has a habit of
accumulating changes due to replication errors at nucleotide positions that
result in little or no effective change in the resulting unit (e.g. tRNA,
rRNA, protein). This sort of change can be so regular as to constitute a
"molecular clock," and is the basis of the concept of Neutral Evolution.
There are a lot of such positions... in protein-coding genes, change in the
third, and to a lesser extent the first, nucleotide in each codon does not
always result in a change in the resultant amino acid (a "silent
substitution"). I know less about this, but I am fairly certain that some
amino acid substitutions do not result in a significant functional
differences in the resulting protein. In functional RNAs, such as tRNA and
rRNA, the actual RNA molecule is a structural unit, and in some nucleotide
positions all that is necessary is a volunteer A, C, G, or U to "fill the
space." Because a nucleotide can only change to three other possible
nucleotides, "convergence" at a single nucleotide position is actually not
unexpected, we recognize that change is likely. Based on this, you can start
to understand a theoretically basis for long-branch attraction.

Having considered that, remember that the "end product" of a gene, a protein
or RNA, is also a *functional* unit, whose morphology is directly controlled
by the DNA nucleotide sequence. Not only are the obviously functional
portions (the tRNA anticodon and amino acids in a protein) coded, but
secondary, tertiary, and quaternary structure are controlled by the
constituent parts of the encoded unit. Thus, a particular nucleotide change,
insertion, or deletion at a particular position in the sequence may produce
a more optimal structure with a functional advantage (morphology!). We
expect that the same "error" might very well occur in two non-reticulating
lineages, and may become fixed in each over time. We would then have
molecular convergence, both at the level of the gene sequence, and at the
level of the product of the sequence. This should not be so difficult to
grasp... it is essentially a special case of the convergence we are used to
thinking about. I don't normally wonder whether convergence I observe
extends to the gene sequence level or not; I suspect most of it doesn't
(even if it did once, the trace of it may have been erased by subsequent
molecular change, but that's another story).

Then, of course, there are *numerous* issues involved in using molecular
data to reconstruct phylogeny. Because sequences can and do change where
they can get away with it ("silent substitutions," loop regions in rRNA,
etc.), particular genes often have a "window" or phylogenetic utility, from
a level at which sequences are so similar as to provide no information, to
one at which either such changes have erased all evidence of the original
path of change, or have altered the gene so much that it is not possible to
determine homology of the individual nucleotide positions. This "window" can
be expressed as a particular level in the taxonomic hierarchy (e.g., the
"family" level... ugh!), or as a time frame (the word on the street is that
mitochondrial rRNA is best of divergences that occurred in the Tertiary).
Use the "wrong gene," and you can get a weird answer... I suspect (with some
degree of confirmation) that this is the source of many of the infamous
"morphology vs. molecules" cases, such as dispute over the highest-level
phylogeny of Aves (see, dinosaurs!) and Amphibia, and molecular support of
the "Haemothermia" concept.

Also, because of the remarkable tendency for animals to go extinct, there
are critical gaps in our knowledge of sequences, gaps that may result in
substantial misinterpretations regarding the ancestral character states
(roughly = sequence) for many groups. This gets worse at more inclusive
levels... we just don't have any non-avian, non-crocodylian archosaurs, no
extant non-paleognathous pan-paleognaths, no non-monotreme non-therian
mammals, etc., until we get to the really nasty parts of phylogeny where we
have a handful of lungfish (often highly divergent, genetically), 1-2
coelacanths, and two robust clades of "fish" with nothing to fill in the
gaps. I suspect this may be the source of the problem in determining the
proper position of Gavialis in Crocodylia, and it probably also relates to
problems in higher systematics of birds... it is certainly something to
watch out for in mammal and general vertebrate phylogeny.

And, a personal observation, often in molecular studies, a single outgroup,
or at best a few, are included to "root the tree." Outgroup character
distributions can affect tree topology; by not providing a broader sampling
of outgroup taxa, you may allow "long branch" problems to effectively
"reroot" the ingroup tree. there are other issues with molecular data, but I
really have other commitments to attend to.

So, molecular data is not (nor is it usually touted as) a panacea. Further,
molecular data are by no means immune from convergence, and any statement
regarding the increased susceptibility of morphology or molecules to
homoplaisy is, IMHO, fairly subjective and unsupportable.

Note that the above is deliberately simplistic, and suffers from my
relatively incomplete understanding of the processes of molecular evolution.
Further, it is written from my (comp. anat./ morphology/ parsimony)
perspective, and will probably sound especially stilted and weird to the
real molecular evolution crowd.

Enjoy, and rip away!

Wagner


Jonathan R. Wagner
2625-B Alcott Lane
Austin, TX 78748