[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Subject Index][Author Index]

Re: Retroposon evidence for an explosive radiation of Neoaves

> I don't pretend to be able to follow everything in this
> paper, but I am a bit puzzled that the tree 
> they present differs in many ways from the "new" views that
> (for instance) falcons are close to Passeriformes.  The
> authors don't comment on the falcon-passerine relationship
> point at all in the text - can anyone explain this to me?

Well if you use *sequence* data, the relative position of falconids and 
passeriforms is the most contentious bit within "higher land birds" (except 
perhaps for the relative position of accipitrids and psittaciforms).

E.g. Fbg intron5 optimizes towards monophyletic "Falconiformes s.l.", 
Crystallin a towards an passerine-accipitrid-strigiform "clade", Aldolase b 
towards passerines as sister to falconids etc. In each case with support values 
of 0.7-0.9, yet usually dangerously unstable with regards to adding/removing 
"rogue" taxa.

Until now, only crystallin a and aldolase b are fairly resistant to "rogues", 
with only _Cariama_ and _Leptosomus_ (former) and Eurypyga_/_Rhynochetos_[*] 
(latter) addition/removal influencing tree topology notably if you use a "core 
higher landbirds" taxon set. But the two loci optimize towards different (and 
in both cases "well-supported") phylogenetic hypotheses.

Altogether, the most widely-supported hypothesis *among all loci* I've tested 
yet is some variant of an "eufalconimorphoid" clade (falcons, passerines, 
psittaciforms, + _Cariama_ and/or coliiforms and perhaps trogons and/or 
"non-gruiform "Gruiformes" in varying arrangements). 
BUT support is neither really strong nor robust (except for aldo-b, where it is 
at least robust); the best-supported and most "rogue-resilient" hypothesis I 
found yet *from a single locus* (crystallin a) places accipitrids and strigids 
closer to passerines than to falcons/psittaciforms.

So it's a bit premature to consider this question resolved.

In any case, I am suspicious about _Cariama_. We cannot get a taxon sample to 
refute LBA/misrooting (there is IIRC no published _Chunga_ sequence, and it 
wouldn't help anyway). All I can say at the moment is that cariamids are 
*probably* not gruiforms.

We NEED the fossil data for "Cariamiformes" and "Leptosomatiformes" included 
before we can draw anything reliable from the molecular data.

And as regards retroposons, considering the amount of homoplasy in the 
"Pegasoferae" case was high (at least 35% IIRC), we NEED to analyze flanking 
sequences and we NEED URGENTLY an intraspecific full-genome study (chicken, 
possibly mallard/domestic duck) to get a fix on how mobile mobile elements are 
Without these analyses one cannot *know* whether mere presence of a transposon 
(any kind of transposon) at an unspecified locus is an apomorphy, or a 
homoplasy. We only know a taxon has it, but it may be highly mobile *within* 
that taxon.

At present the assumptions of transposing frequency in birds are simply 
guesstimates not supported by any hard data. I had the odd transposon (all 
sorts) in my taxon/locus set, and usually they were devoid of any pyhlogenetic 
pattern, or almost so. I.e. they may "jump" more quickly than generally 

In any case, the claim that "one should not BLAST retroposons" is dangerously 
untrue, at least according to the (plant) retroposon experts I asked.
You need to be careful with your BLAST parameters though. Using the standard 
settings will indeed yield only nonsense. But if you account for partial 
sequence loss/gain, you will get *interesting* results... you start to wonder 
how exactly these things originate, and what their role is in evolution (the 
"junkyard tornado" may be more appropriate a metaphor than usually assumed).

In short, retroposons are interesting, but we know too little of them (and most 
of what we *know* is from plants) to place much weight on them at present. 
Certainly, the experience in botany is mixed... (the "Eufalconimorphae" paper 
only cited one of the several papers in which reservations against retroposons 
are discussed, and it didn't *discuss* it at all).
And I seriously wonder whether the quest for the holy grail of "homoplasy-free" 
characters is not a blind alley.
Perhaps improving the (initial) *qualitative* analyses is the actual "best" way 
to improve the (subseqent) quantitative analyses. That approach will allow to 
more reliably detect homoplasy, which is the way to go if we can't get rid of 
it (which may be the case).

Because from my own analyses I find that even a state-of-the-art cladistic 
analysis can be fooled (entirely incompatible phylogenetic hypotheses *at 
"sufficiently" high support* are something that *should* not happen), and *the 
more sophisticated technologies for signal recovery you use, the more easily 
can it be fooled* (because you make *more* assumptions unsupported by actual 
factual evience).



* In my first msg regarding the new paper, I hadn't tested these yet for this 
locus. And as it turned out, they are indeed the only taxa in my set whose 
addition/removal *does* have a strong effect on tree topology here.

FWIW, I have to date found only one (apparently) unequivocal indel apomorphy 
within "higher landbirds" that is nontrivial in its phylogenetic signal (i.e. 
not a simple autapomorphy of taxa we *knew* to be related anyway). And even 
point-mutation ?apomorphies with good signal are rare (point mutations are of 
course prone to homoplasies).

That one apomorphy may be good, though. My "list of core higher landbirds" was 
based on it. But it may still be homoplasious between "non-gruiform 
'gruiforms'" and "higher landbirds" (it is deleted in both).

Next week I'll be back at work and look at some more data. Til then, I might 
still run the occasional analysis from my laptop, but don't expect much.