Re: How are columbiformes (doves and pigeons) related to psittaciformes?

[Christopher Taylor <ck.taylor@auckland.ac.nz>]

Dear Dora,

    I'm afraid that statistics isn't my strong point, but I'll answer as
best I can. Prepare for Molecular Evolution -101...
    The problem is that genetic change doesn't appear to always be constant.
A lot of older molecular phylogenies (and a few recent ones - grrrr...) were
calculated under the assumption that it was, as a result of the Neutral
Theory of Molecular Evolution. Basically, this says that the majority of
molecular mutations have no effect on the selective fitness of the organism
(the genetic code - how DNA is transcribed into proteins - is redundant, so
many changes don't result in a change in the final product. Also, a large
part of your DNA is actually seemingly useless junk which doesn't actually
do anything, it just sits around taking up space). In the absence of
selection, mutation just happens at random, and over time, the rate would
appear as near to constant as makes no odds.
    But this is not entirely the case. Selection does appear to make
molecular change happen faster or slower, depending on the selective regime.
The background rate of evolution is still random (as far as we know), but
the chance of a mutation becoming widespread in the population is affected
by selection.
    It depends on the molecule, too. For the most part, functional DNA is
affected by selection, the aforementioned pieces of useless junk are
probably not.
    There is also a problem with saturation. DNA is made up of only four
bases, so the chance of convergence is fairly high (if an A changes to a G
in one organism, and to a T then a G in another, there's no way to tell from
the final result that the two organisms actually had different evolutionary
histories - a G is a G).
    These factors can lead to what is generally referred to as Long Branch
Attraction, where two widely separated branches on the tree can be joined by
convergent evolution (the same sort of thing can happen in morphological
phylogeny - as the number of autapomorphies in a taxon increase, the chance
becomes higher that they're hiding or replacing the synapomorphies with
related taxa).
    Fortunately, methods have been developed (and are continually being
improved) to identify saturation, and compensate for different evolutionary
rates. Basically, they run on the principle that we can identify what kinds
of molecular mutations are more likely to occur than others, so changes
which happen often can be accorded less importance in calculating the
phylogeny, while changes that happen rarely can be accorded more importance.
One advantage of molecular phylogeny is that, because we are dealing with a
small set of chemical parameters, it's much easier to calculate such
starting biases.
    Of the methods for identifying saturation, one which I know of is based
on two types of mutation called transitions and transversions. The four
bases fall into two pairs chemically, A and T vs. C and G. A mutation which
changes one member of a pair to the other is a transition, one which changes
a bases to a member of the other pair is a transversion. Chemistry indicates
that a transition can happen more easily than a transversion, so if the
sequences are not saturated, there should be a lot more transitions than
transversions. If the number of transversions is too close to the number of
transitions, it implies that your sequences are probably saturated, and your
data is just so much random junk.
    The Passeriformes case appears to be a perfect case of Long Branch
Attraction. When I mentioned that I found a basal Passeriformes in my
amateur analysis, I was alerting people's attention to the fact that I'd
possibly not done enough to test the integrity of my data against
saturation, etc., and so my results should be held at arm's length and
inspected closely, like an old bottle of wine that might be rather good, or
might have just turned to vinegar. In earlier molecular phylogenies of
birds, Passeriformes appeared to be quite basal, with strong support.
However, as methods of molecular analysis have improved, the Passeriformes
have moved up the tree - their earlier low position appears to be the result
of particularly fast evolution resulting in a lot of unique features that
drown out those they share with other taxa. We still have pretty much no
idea where Passeriformes _do_ fit - the misleading data is still playing
games with us :) - but we're on the way.
    Hope all this helps a little,

        Cheers,

            Christopher Taylor

On 13/5/04 11:13 pm, "Dora Smith" <villandra@austin.rr.com> wrote:

> Christopher:
> 
> I am wondering if method of statistical analysis could be contributing to
> this difficulty in resolving the main direction of bird evolution - if it
> was backwards (passeriformes) or forward (from primitive birds).
> 
> If I were going to do a project of this sort, I would look chart differences
> in genetic coding - and leave it at that.
> 
> My own background is all but degree at the Masters' level in social science,
> with an associates' in science.   I concentrated on biology and chemistry in
> my first two years of college, and studied alot of statistics in graduate
> school.
> 
> Some people are doing statistical methodologies that I don't follow at all,
> and some are reporting log odds ratios and appearing to use them as
> regression coefficients, which suggests they are doing log linear analysis
> in an effort to demonstrate that their findings achieve statistical
> significance.
> 
> That requires that one form hypotheses about directionality of what came
> before what or caused what - and there are any number of ways that
> regression coefficients or log odds ratios could cause two phenemenon that
> are related to have backwards directionality or spurious directionality.
> 
> Statistical significance is nice, but it will never beat logic.  I would
> really just construct tables showing the actual genetic changes across bird
> orders and leave it at that.   Common sense tells you that there will be
> fewer genetic differences between closely related species, and more genetic
> differences between distantly related species, and that some genetic trait
> should be common to all of them.    One ought to be able to follow the
> accumuluating differences to what birds have been around longest, if indeed
> any of them have.  ...  Oh well, don't make me go through and get the genes
> from Genbank or wherever and do it!
> 
> If genes have evolved in parallel ways through convergence, then one ought
> to be able to demonstrate why that would happen by pointing out what the
> gene does that is so strongly adaptive that it would evolve independently
> more than once.   In the articles I have read, of which there are maybe ten,
> noone has discussed what a single piece of the genetic code they examine
> actually does!
> 
> Would mitochondrial DNA in particular be likely to evolve independently in
> similar ways through convergence?    It kind of seems unlikely - unless it
> had something to do with the energy needs of the animal in question!
> Could birds do the whole oxygen and metabolic thing so differently that they
> need different mitochondria to do it?
> 
> Anyhow, exactly what method of statistical analysis did you use to arrive at
> your conclusion that passeriformes are basal to birds?
> 
> Yours,
> Dora Smith
> Austin, Texas
> villandra@austin.rr.com
> ----- Original Message -----
> From: "Christopher Taylor" <ck.taylor@auckland.ac.nz>
> To: <dinosaur@usc.edu>
> Sent: Monday, May 10, 2004 5:57 PM
> Subject: Re: How are columbiformes (doves and pigeons) related
> topsittaciformes?
> 
> 
>>     Interestingly enough, I did a project for an MSc paper a few years
> back,
>> where I attempted to construct a phylogeny of Aves using 12S and 16S rDNA
> to
>> test the monophyly or otherwise of Pelecaniformes. The final tree was
> almost
>> totally unresolved, but one of the very few well-supported clades which
> did
>> appear was Psittacidae + Coliidae. Unfortunately, it wasn't the main focus
>> of my project, so I didn't look at it further. (I also found strong
> support
>> for Passeriformes basal to Neoaves - you have been warned.)
> 
>