Cau's 2018 theropod tree

Hi all.  Work's been rapid on the Ornithoscelida project, which has improved due to David Marjanovic's request that I provide a citation for each character state we think was inaccurately coded by Baron et al..  It's time consuming, but has revealed several times where I was mistaken in my scoring, and forces you to really evaluate whether you think the evidence is good enough to truly call someone's character state choice wrong.  As tedious as it is, this should really become the standard for scoring justification in any analysis (and no, I didn't do that for the Lori analysis, though I have the preliminary steps for a far future offshoot of it).

Speaking of the Ornithoscelida paper, coauthor and fellow theropod blogger Andrea Cau just got a big new paper published using his MegaMatrix character list.  So congrats!  It's a summary of theropod evolution focusing on the path from Teleocrater to Meleagris, detailing the changes in each node from Pan-Aves (well, technically that's a stem) to Aves, and using that data to examine the trends and rates of change along that line.  We get two new clade names as well.  Dracohors is (Megalosaurus bucklandii < - Marasuchus lilloensis), so covers silesaurs and dinosaurs.  The use of a stem instead of a node like Silesaurus+Megalosaurus was wise considering some analyses (especially those recovering ornithischian silesaurids) recover Lewisuchus and such at an uncertain position relative to Silesauridae, but it's always closer to dinosaurs than Marasuchus.  But come on, 'Dracohors'?  Dragonian prostitutes?  Did no one learn from Ascendonanus?!  Oh, and the plural is 'dracohorsians', which sounds like a hybrid race from My Little Pony.  Maniraptoromorpha is (Vultur gryphus < - Tyrannosaurus rex) and is a sorely needed name that I could have used in the Lori paper, since we ended up limiting the taxon sample to maniraptoromorphs.  Recall that the results of TWG characters get worse as you get further from Norell et al.'s (2001) or Clarke's (2002) original scopes, and I didn't want to be Petersian and tackle e.g. the Megaraptora problem without the Carrano et al. (2012) or Novas et al. (2013) characters while using a batch of maniraptoromorph characters.  The name Maniraptoromorpha also fits the -morpha > -formes > basic clade pattern, which I like.

Sacrum of the maniraptoromorph Bambiraptor feinbergi (holotype AMNH 30556) in ventral view, anterior to left (courtesy of the AMNH).

Cau's analysis itself has positives and negatives.  What everyone knows about the MegaMatrix is that it's HUGE.  I'm talking 1781 characters.  That's more than all Mesozoic theropod analyses except the poorly coded Livezey and Zusi (2007), which used 2954 characters and focused on living birds though technically included 30 non-avian theropods too.  But if you look at the matrix, the size is a tiny bit misleading because Cau doesn't use any multistate characters.  So instead of a single character of "Premaxilla - number of teeth - three or less (0); four (1); five (2); six (3); seven or more (4)", he has-
"14): Premaxilla, fifth alveolus: absent (0); present (1). ...
1717): Premaxilla, sixth alveolus: absent (0); present (1).
1718): Premaxilla, seventh alveolus: absent (0); present (1).
1759): Premaxilla, fourth alveolus: absent (0); present (1)."
Or instead of "Sacral vertebrae - number - two or less (0); three (1); four (2); five (3); six (4); seven (5); eight (6); nine (7); ten (8); eleven or more (9)", he has-
"343): Third sacral vertebra: absent (0); present (1). ...
1707): Fourth sacral vertebra: absent (0); present (1).
1708): Fifth sacral vertebra: absent (0); present (1).
1709): Sixth sacral vertebra: absent (0); present (1).
1710): Seventh sacral vertebra: absent (0); present (1).
1711): Eight sacral vertebra: absent (0); present (1).
1712): Ninth sacral vertebra: absent (0); present (1).
1713): Tenth sacral vertebra: absent (0); present (1).
1714): Eleventh sacral vertebra: absent (0); present (1)."
Now this is largely fine, as both ordered and unordered multistate characters can generally be scored this way using certain procedures to avoid weighting.  Cau does it to make Bayesian analyses easier (pers. comm.).  But the Lori character list would be 25% longer (875 non-zero states in my 700 characters) using this method, for instance.  Of course the MegaMatrix was designed for a larger taxonomic scope, so will necessarily have more characters, those meant for carnosaurs, ceratosaurs, basal dinosauromorph phylogeny, etc..  Limiting the taxon sample to maniraptoromorphs finds 1170 parsimony-informative characters in Cau's dataset.  There are also 14 parsimony-uninformative characters in the Lori dataset, since some TWG characters were designed for tyrannosauroids or positioning outgroups like Dilophosaurus, Monolophosaurus, etc.. This would leave 861 Cau-style characters in the Lori analysis, so 74% as many as the MegaMatrix.  Not so bad...

It also doesn't use any polymorphic scoring, so there are no 0+1 scores in Cau's matrix.  I'm not sure if this matters analytically, since I assume he scores a polymorphic taxon unknown, and I don't know how TNT treats polymorphies when it's making trees.  Not that the standard method is without its problems, since TNT is stupider than PAUP in treating uncertainty polymorphies (could have either four or five teeth) the same as actual polymorphies (some have four teeth, others five).  He also doesn't differentiate inapplicable (-) and unknown (?) scorings, which is fine functionally as I believe TNT treats them the same, but this and the lack of polymorphic scores do make the matrix itself more opaque to users since a lot of question marks there aren't technically unknown.  For the Lori matrix, we're providing both a TNT and a NEXUS file, with the latter showing uncertainty polymorphies vs. actual polymorphies, plus when taxa can but shouldn't be scored due to ontogeny (e.g. a juvenile with unfused bones, when adults are unknown).

Cranial elements of the maniraptoromorph Microvenator celer (holotype AMNH 3041), with supposed right lacrimal of Makovicky and Sues in lateral view (upper left, anterior to right), dentary in dorsal view (lower right, anterior to top) and unidentified elements (scale bar in cm) (courtesy of the AMNH).

The taxon sample is interesting because it is very reduced compared to the Halszkaraptor analysis (Cau et al., 2017) that used the same character list (plus eight new characters; though among maniraptoromorphs, Jianianhualong and Chongmingia were added).  Alvarezsaurids are only represented by Alvarezsaurus and Patagonykus, caenagnathoids by Anzu and Khaan, therizinosaurs by Beipiaosaurus and Falcarius (and Jianchangosaurus... see below), etc..  Which still leaves it as the best large scale Mesozoic theropod analysis published, but without e.g. undisputed alvarezsaur skulls (since Haplocheirus finds itself elsewhere in some analyses), it's bound to get things wrong.  And it does differ from Cau et al.'s 2017 trees which focused on maniraptoromorphs.  Zuolong is outside Neotetanurae instead of a coelurosaur, Coelurus is a maniraptoromorph [see how useful this clade name is!] instead of a tyrannosauroid, Jianchangosaurus is an alvarezsaur(!), therizinosaurs are caenagnathiforms instead of outside Pennaraptora, Fukuivenator is a paravian instead of a therizinosaur, halszkaraptorines are unenlagiines, Xiaotingia is a scansoriopterygid instead of an anchiornithid, scansoriopterygids are further from Aves than Archaeopteryx, and Zhongjianornis is in Fake Ornithuromorpha instead of sister to Pygostylia.  Cau states on page 4 that taxon choice was based on a subjective (or at least not objectively justified) "balanced series of criteria" and I get that the point of the paper is trends leading to Aves, but having the quite complete Jianchangosaurus as the first branching of four alvarezsaurs, and eliminating a major step by placing therizinosaurs sister to oviraptorosaurs, seem like problems.  I'm sure the main trends are only very minorly affected, but still.  I think if I needed a reduced taxon sample (which was a real concern during part of the Lori process...) and was being subjective anyway, I would fiddle with the sample until I got a topology that matched the full analysis.  Or maybe Jianianhualong, Chongmingia and/or the added eight characters changed the topology, and this whole paragraph is misled.  Or maybe the reduced outgroup in the Halszkaraptor matrix affected the topology (only eight non-maniraptoromorphs compared to 58 here), which would be problematic for coelurosaur analyses since most have just been using 2-5 of the same outgroup taxa (Sinraptor dongi, Allosaurus, Monolophosaurus, Dilophosaurus, and/or Coelophysis...)..

Parsimony-based strict consensus of Cau (2018)with clade names along the bird lineage and Decay Indices > 1 (after Cau, 2018).

What of the phylogenetic results themselves?  Eodromaeus and herrerasaurs are in an unresolved polytomy with sauropodomorphs and ornithoscelidans, which matches their increasingly uncertain position now that people are using large taxon samples for more than Nesbitt-derived matrices and Sereno's non-testing 'analyses' from the 90s are fading into history.  Ornithoscelida itself is recovered, but with only Heterodontosaurus and Tianyulong representing Ornithischia, I don't give that much credit regardless of the character count.  Heterodontosaurids are very theropody in some ways few other ornithischians are and generally fall out as the first branching clade in analyses (now that Pisanosaurus is a silesaurid, which Cau also recovers) and ignoring Chilesaurus for the moment.  But I'm co-writing a whole paper on that, so let's move on to theropods.  Cryolophosaurus is a dilophosaurid coelophysoid, which is counter the recent consensus of finding it closer to averostrans or even a tetanurine itself.  As stated above, Zuolong is outside Neotetanurae, which is weird since it generally falls out as a basal coelurosaur.  Chilesaurus is also here, matching its authors' conclusion, but I think the poor ornithischian sample keeps this analysis from adequately dismissing an ornithischian identity.  But I'm co-writing a paper on that and etc. etc..  Back to theropods.  Cau recovers a Rauhut-like Carnosauria including megalosaurids, piatnitzkysaurids, Monolophosaurus and allosaurs.  Neovenator is sister to Allosaurus and Sinraptor sister to Acrocanthosaurus among the latter, which harkens back to late 90s phylogenies.  Megaraptorans (including Gualicho) are tyrannosauroids closer to Tyrannosaurus than Eotyrannus, matching Novas et al. (2013).  The tree is untraditional in placing compsognathid-grade taxa outside Tyrannoraptora, though Novas had that opinion back in the early 1990s.  As stated above, the maniraptoran section has the out of vogue caenagnathiform therizinosaurs and the very weird alvarezsauroid Jianchangosaurus, but those might be glitches due to the taxon sample.  Fukuivenator is sometimes a dromaeosaurid and sometimes the basalmost paravian, which its authors recovered with the addition of three steps.  Cau finds avialan troodontids, standard for his published analyses.  Like his Halszkaraptor version, Rahonavis is close to jeholornithids and Protopteryx can be outside Ornithothoraces (it always is in the 2017 trees).  The latter is funny because I had that idea way back in 2000 on the DML ("My analysis of 31 characters and 6 taxa resulted in a single most parsimonious teee (CI .81, HI .19, RI .78)" ... how quaint), so if it turns out to be true that would be something.  Finally, presumably due to excluding fragmentary taxa like Teviornis, the huge ornithuromorph polytomy of the 2017 paper is resolved and Patagopteryx falls out by Apsaravis, Ichthyornis and Hesperornis+Aves.  Wha?

Interestingly, Cau also performed a Bayesian analysis.  In these, herrerasaurs are sister to Dinosauria like Baron and Williams (2018), but Eodromaeus is a basal ornithoscelidan.  Allosauroid relationships now match the Carrano et al. consensus, and Zuolong is a basal coelurosaur but Chilesaurus follows it.  Gualicho is now closer to tyrannosaurids than to megaraptorans.  Jianchangosaurus is back to being a therizinosaur, but Ornitholestes and Haplocheirus are basal ornithomimosaurs.  Weird.  Therizinosaurs are no longer sister to oviraptorosaurs, and actually move down an extra node to be stemward of alvarezsaurids too.  Especially interesting is that scansoriopterygids move way down from being avialans to being basal oviraptorosaurs.  Are they the missing Jurassic oviraptorosaurs?  This lets Xiaotingia go back to Anchiornithidae.  Rahonavis is now sister to Jeholornithidae instead of a couple nodes stemward of it, and Sapeornis is a confuciusornithiform.  Protopteryx is always an enantiornithine (aw...), but Patagopteryx is still way crownward, though now Vorona joins it (basal fake ornithuromorph in the parsimony analysis).  Well, that does match their Late Cretaceous age at least.  Considering all of these changes, I like a few things better about the parsimony version and a few things better about the Bayesian version.

How does Cau's Maniraptoromorpha topology compare to Lori's?  Quite different on both a broad level and with regard to the detailed relationships within the clades, although there are a few things that aren't common but popped up in both.  I'm dying to say more, but that post has got to wait.  Congrats again to Andrea.

References- Norell, Clark and Makovicky, 2001. Phylogenetic relationships among coelurosaurian theropods. In Gauthier and Gall (eds.). New Perspectives on the Origin and Early Evolution of Birds: Proceedings of the International Symposium in Honor of John H. Ostrom. 49-67.

Clarke, 2002. The morphology and systematic position of Ichthyornis Marsh and the phylogenetic relationships of basal Ornithurae. PhD thesis, Yale University. 532 pp.

Livezey and Zusi, 2007. Higher-order phylogeny of modern birds (Theropoda, Aves: Neornithes) based on comparative anatomy. II. Analysis and discussion. Zoological Journal of the Linnean Society. 149. 1-95.

Carrano, Benson and Sampson, 2012. The phylogeny of Tetanurae (Dinosauria: Theropoda). Journal of Systematic Palaeontology. 10(2), 211-300.

Novas, Agnolin, Ezcurra, Porfiri and Canale, 2013. Evolution of the carnivorous dinosaurs during the Cretaceous: The evidence from Patagonia. Cretaceous Research. 45, 174-215.

Cau, Beyrand, Voeten, Fernandez, Tafforeau, Stein, Barsbold, Tsogtbaatar, Currie and Godefroit, 2017. Synchrotron scanning reveals amphibious ecomorphology in a new clade of bird-like dinosaurs. Nature. 552, 395-399.

Baron and Williams, 2018. A re-evaluation of the enigmatic dinosauriform Caseosaurus crosbyensis from the Late Triassic of Texas, USA and its implications for early dinosaur evolution. Acta Palaeontologica Polonica. 63(1), 129-145.

Cau, 2018. The assembly of the avian body plan: A 160-million-year long process. Bollettino della Società Paleontologica Italiana. 57(1), 1-25.