Re: Mesozoic roots of parrots and passerine birds

[evelyn sobielski <koreke77@yahoo.de>]

> This paper also propose the
> split between falcons and hawks:
> 
> A Phylogenomic Study of Birds Reveals Their Evolutionary
> History
> Shannon J. Hackett, et al. Science 320, 1763
> (2008); DOI:
> 10.1126/science.1157704

This and the others cited all use beta-fibrinogen intron 7 data, which is 
suspected to carry a strong bogus phylogenetic signal. They cannot be taken at 
face value, until the data have been reanalyzed. 

See here: http://www.biomedcentral.com/1471-2148/8/20

(Give me some 15 more months, and I'll probably have done that ;-) )

Same problem as with transposons: indels were once believed to be entirely 
nonhomoplasious, but they are apparently anything but that. I don't know which 
it was, but one mammal study found a signal-noise ratio of 1.5 among about a 
dozen indels.

If you look at Supplement 1 of the present paper, you'll find that support for 
Eufalconimorphae rests mainly on the J-class retroposons, and that comparison 
across taxa is otherwise highly incomplete. It looks as if they took the 
retroposon set of the zebrafinch and sequenced other birds until they had 
enough data to "prove" Eufalconimorphae. But verificationism hasn't been an 
acceptable scientific methodology 100 years ago! You need to look for data 
which REFUTE your hypothesis, and if you can find none, THEN you're cool. (It 
is also interesting that their data does not support monophyly of either 
Cypselomorphae and Mirandornithes)

The bottom line is that non-sequence molecular data presents similar problems 
as sequence data turned out to be 20 years ago (see e.g. 
http://www.springerlink.com/content/1l5c766g0cptp6fv/). But while we have by 
now worked out the source of bias and bad SNR in sequence data, the use of 
non-sequence molecular data is too recent to have done so. In 10 years, we can 
use such markers and really yield robust results, but until then it is mnrely 
clear that these genomic events are not as random as they seem. So one *can* of 
course use the data, but "robust" and "definite" are terms one should not use 
to describe the results. Except if the SNR is 0 (i.e. the data are just random 
noise) - and possibly even then -, today's phylogenetic analyses WILL ALWAYS 
resolve a phylogeny. 

See e.g. http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2577343/ for some of the 
current research in this direction. This is about a comparatively 
well-understood transposon, yet it took 10 years or more to clarify at least 
the basics of its target site preference.

In short, until J-class retroposons are proven to insert at random sites, these 
results are evidence but not proof.


Regards,

Eike