Re: New paper on Neoaves

[David Marjanovic <david.marjanovic@gmx.at>]

> One question to keep in mind is: Do the analyses that incorporate single 
> gene regions yield the same or similar topology as the combined dataset, 
> but with less support; or do they yield a very different topology to the 
> combined dataset tree, with strong support?  In other words, does the tree 
> based on a combined dataset (=5007 bp drawn from five different gene 
> regions) improve the resolution of each of the single-gene trees, or does 
> it contradict these single-gene trees?  I'd hate to think that there is a 
> belief that you can 'improve' the phylogenetic signal simply by expanding 
> the dataset (=stringing together more and more DNA sequences).

Why? Of course you can. More data can never hurt, and as long as you haven't 
reached a certain (high) limit of characters per taxon, it actually adds 
signal. The idea is that the signal adds up while the noise cancels itself 
out. Why should the noise conspire to create a false signal?

With molecular data, adding more and more genes (or regions of the same 
gene) has an additional advantage: they won't evolve at the same speed. Some 
genes will evolve fast and thus will be able to resolve recent divergences, 
while saturation effects like long-branch attraction and long-branch 
repulsion will abound for deeper splits; other genes will evolve more 
slowly, resolving deep divergences but failing to find the recent ones 
because they don't have enough apomorphies yet. Ideally, each of the genes 
in the dataset will have its own rate of evolution, and because noise 
cancels out rather than adding up, you will get a fully resolved tree.

Clearly, the new Neoaves analysis is not ideal. But what are 5007 bp? I'm 
currently writing about one that used the entire mitochondrial genome --  
11069 bp, of which some 6000 are parsimony-informative IIRC -- and still 
produces a suspect clade whose existence should be tested by better taxon 
sampling.

> Also, are each of the single-gene trees consistent?  Individual gene 
> regions
> may be better at capturing divergence at a certain level, which means that
> you're going to get a very muddled phylogenetic signal if you simply 
> string
> these genes together.

If there is a signal, it will add up. The noise will cancel itself out. The 
only way that I can think of to get a fake signal is by having a skewed base 
frequency distribution, and outside of parsimony, this is among the first 
things analyses take into account these days.

> As a general principle, does increased taxon sampling even help?  I'm not 
> certain it does.  If a gene is unable to capture the phylogenetic signal, 
> or the analysis is unable to tease out this signal, no amount of taxa will 
> improve the situation.

In that case better taxon sampling will at least show that the gene really 
is incapable of resolving divergences that happened within the time frame in 
question. :-)