There are some apparent problems in the published mitochondrial data (sequences of the same gene from different studies placing the same species in completely different positions; sequences of two mitochondrial genes placing the same specimen at completely different positions, etc.), but nothing obvious in the sequences I used here, to the exception of a somewhat suspect high variability in the cytb sequences of Sylvia borin -- it may be that some of the published sequences are a bit off, but I can't easily determine which ones.
There are two distinct mtDNA data set that have a broad taxonomic sampling of
Sylvia spp. in GenBank:
1) One, with cytochrome-b sequences only, associated by Katrin Böhning-Gaese's papers [
here]. (Most of these were actually by Andreas Helbig. This data set was also used to build the phylogenetic hypothesis suggested in Shirihai et al's
Sylvia Warblers.)
2) One, with cytochrome-b and ND2 sequences, deposited by Gary Voelker and co-authors [
here].
There are two significant problems of incongruence in these data, that would be expected to affect the results of a phylogenetic analysis.
1) Based on cyt-b sequences (of
both data sets),
S. deserticola is part of the
S. undata complex,
S. mystacea is sister to
S. melanocephala, and
S. melanothorax is closest to
S. ruppeli. (This also makes some sense morphologically.) In sharp contrast, the ND2 sequences of
S. deserticola,
S. mystacea and
S. melanothorax are very similar (divergence might be consistent with the three sequences being from the same taxon), and place all three species together in a rather basal position. (This makes basically no sense morphologically.) This is reflected in
Voelker & Light's published tree where these three species form a clade; but, in this tree, the three species don't appear as similar as with ND2 only (because their cyt-b sequences actually differ a lot), and they appear closer to the core-
Melizophilus group than with ND2 data only (because their cyt-b sequences all belong within this group; it is also likely that the rooting of the core-
Melizophilus group was affected by the inclusion of these data). No such relationships are suggested by any other data. For
S. melanothorax, another ND2 sequence is available from Price et al 2014 [
here], which differs strongly from that of Voelker & Light and places the species in a position fully consistent with that suggested by cyt-b data. For
S. mystacea, several cox1 sequences are available [
here]: one of these is an obviously misidentified
S. communis, but the other two congruently make the species the sister group of
S. melanocephala, just like cyt-b data. Because they are in strong conflict with all other available data, I believe these three ND2 sequences to be most likely incorrect. (But I admittedly can't figure what these sequences are likely to have really been.)
2) There is a problem of incongruence in the cytochrome-b sequences in the
S. undata - deserticola - balearica - sarda complex. The 8 cyt-b sequences from this complex (one for each species in each data set) all clade together but, within this clade, they do not cluster at all according to the species to which they are attributed. Only the two
sarda sequences are reasonably similar to one another; the three other Voelker & Light sequences are closer to them than
any of the three other Böhning-Gaese/Helbig sequences; the three other Böhning-Gaese/Helbig sequences form an apparent sister group to the other five sequences. I don't believe that these sequences can all be correct, but it's quite hard to say for sure which are and which are not. However, based on the Böhning-Gaese/Helbig cyt-b sequences,
sarda appears basal in this complex, while, based on the Voelker & Light cyt-b sequences,
undata appear basal; the former is consistent with what ND2 sequences suggest, the latter is not.
I used mtDNA (ND2, cyt-b, COI) and nuclear (βFib5, GAPDH, MUSK, MYO, ODC, RAG1 and TGFβ2) data from GenBank to create a supermatrix encompassing Sylviidae species (
sensu Clements), plus representatives of 6 genera of Paradoxornithidae. Given the above notes, I discarded the three "suspect" Voelker & Light ND2 sequences, and chose to use the Böhning-Gaese/Helbig cyt-b sequences for the
S. undata group. I also discarded the species that would have been represented in this matrix by a cyt-b sequence only (often not enough for confident placement, and this would have resulted in these species' sequences having zero overlap with the sequences of some other species, which I generally prefer to avoid). (Note that, currently,
S. hortensis is almost certainly represented in published data by its Böhning-Gaese/Helbig cyt-b sequence only. Other sequences are attributed to this taxon in GenBank (the cyt-b and ND2 from Voelker & Light, one COI barcode), but these do not differ significantly from sequences of
S. crassirostris and, upon checking, they all appear to be from Greece where of course
S. crassirostris is the expected form.)
I guess I could try to see how support behaves if I add more data.
Well, unfortunately, things did not improve at all. (The signal in some of the added genes I added conflicts with a sister-group relationship of the two
Sylvia.)
I have attached a .pdf file with separate ML consensus trees build from mtDNA and from each nuclear marker, and another one with two ML consensus trees based on a concatenation of all markers (one with sparser taxonomic sampling; the other one with more species included, but many of them represented by mtDNA only).
Not sure about TiFs Curruca and would like to see comparative genetic distances over the entire Sylviidae/Paradoxornithidae clade.
Hoping this will please you, then...
The distance is quite big. At first sight, only
Myzornis vs. all other Paradoxornithidae might really pretend to compare to
Sylvia vs.
Curruca as recognised in TiF.
