RE: New paper on Neoaves

[Christopher Taylor <gerarus@westnet.com.au>]

    To correct something Tim said quite a bit earlier, Ericson et al.
did find Palaeognathae to be 'monophyletic', but this wasn't shown as
such in the paper because Palaeognathae is the outgroup. If you mentally
uproot the published tree, you'll see what I mean.
    It is true that there are systematic biases that can be at work in
molecular data - this is why likelihood methods rather than parsimony
have rapidly become the standard for molecular analysis. There can be
such biases in morphological data, too - witness the determination of
loons and grebes to stick to each other due to characters related to
foot-propelled diving. Molecular data do have an advantage over
morphological data in that it can be easier to recognise and quantify
these biases - at least for the big and obvious ones, such as
transition/transversion ratio. What I am not aware of any studies on so
far is whether any such biases that may be taxon-variable (such as codon
bias) are found throughout the genome of any given species, or whether
different parts of the genome may differ in these biases. I'm not sure
if such a study could be done at the present point in time, as it would
potentially require a large number of fully sequenced genomes for
different taxa (we do have a large number of bacterial genomes, but
factors such as lateral gene transfer muddy the waters somewhat so the
results may not be relevant for eukaryotes). Stringing sequences
together from different genes may obscure differences in local biases,
but to analyse the genes separately and then construct a supertree from
the results may mean that the authors miss any 'hidden' support shared
between genes. As far as I know, no-one has yet come up with a solution
for this problem - one pays one's money, and all that.
    Problems with trees from mitochondrial data can't necessarily be
directly inferred for nuclear data. Mitochondrial genomes are generally
inherited asexually, without any sort of intermixing, and without
recombination. As such, stochastic drift is more likely to have a
significant effect within a mitochondrial lineage. Also, a mitochondrial
genome should be effectively regarded as a single unit, which may not
necessarily be the case with the nuclear genome, as different sections
may have different evolutionary histories.
    Part of the problem with commenting on which characters in molecular
analyses support which clades is that this may not be a simple
one-to-one relationship in likelihood methods. Unlike parsimony, which
calculates the most likely tree given the data, likelihood calculates
what is the tree most likely to give the dataset under the model
assumptions used in the analysis. One effect of this is that, unlike a
parsimony analysis, the tree can't really be taken apart piecemeal and
examined as separate branches. The distribution of characters in one
part of the tree is significant for the results of a separate part of
the tree, because they affect the model parameters that may be
calculated for the analysis.

    Cheers,

        Christopher Taylor

-----Original Message-----
From: owner-DINOSAUR@usc.edu [mailto:owner-DINOSAUR@usc.edu] On Behalf
Of Tim Williams
Sent: Sunday, 13 August 2006 2:41 AM
To: dinosaur@usc.edu
Subject: RE: New paper on Neoaves

Mickey Mortimer wrote:

>I wonder.... how many well supported (>95% Bayesian or bootstrap)
molecular 
>nodes have later been 'disproven'?  And of the ones that have been, in
how 
>many cases was it a matter of needing to throw more taxa or bases in
the 
>analysis?

You and David hold to the assumption that homoplasy is random, and can
be 
overcome simply by enlarging the dataset or expanding the taxon sample.

Thus, the phylogenetic signal, which denotes a common and shared
ancestry, 
should eventually overcome this randomly distributed 'noise'.  But with 
genes and proteins, this is often a brave assumption to make.

Here's one case that highlights the problem.  Naylor and Brown (1998)
used 
over 12,000 bases (from 19 mitochondrial genes) and recovered 100%
bootstrap 
support for a clade that comprised vertebrates and echinoderms to the 
exclusion of amphioxus (lancelet).  This topology was recovered
regardless 
of the method of analysis.  The authors didn't believe their own tree,
since 
it is contradicted by compelling morphological evidence that vertebrates
are 
closer to lancelets than to echinoderms.  They argued *against* simply 
accruing even more taxa or even longer datasets (not possible for the 
mitochondrial genome - they had already used every gene!).  Instead,
they 
favored investigating the underlying factor(s) that are pulling the 
vertebrate and echinoderm sequences together, and dispute the assumption

that homoplasy ('noise') is distributed randomly in the dataset.

I know I'm on my hobby horse here, and I apologize for the verbose
postings. 
  But the thing that is obvious about *morphology*-based phylogenetic 
analyses is that they are almost always followed by a discussion of
which 
morphological characters (synapomorphies) unite which taxa.  In other
words, 
it's plain to see the identity of the characters that diagnose certain 
clades.  This rarely happens with *molecular* clades.  Here, the
characters 
are at the level of genes and amino acids, and the structural and
functional 
properties of the sequences are skimmed over.  Instead, researchers tend
to 
focus on bootstrap support (or posterior probabilities, in the Bayesian 
world) as the final determinant for a 'good' tree, and move on.
However, 
I'd like to see more discussion of the gene- or amino-acid-level factors

that are responsible for the topology of the tree.  For example, what 
sequence-level characters are putting Coronaves together, or pulling 
Falconiformes apart?

Cheers

Tim


-- 
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.10.9/417 - Release Date:
11/08/2006
 

-- 
No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.405 / Virus Database: 268.10.9/417 - Release Date:
11/08/2006