Molecular versus morphological trees for Crocodylia — Salas-Gismondi et al., 2016 as an example

A new fossil crocodylian species was published in PLoS ONE last April[1], Gryposuchus pachakamue. The authors presented an expanded phylogenetic dataset for gavialoids based on previously existing crocodylian datasets but significantly reduced the number of alligatoroids and crocodyloids represented (their justification being that the internal resolution of Brevirostres, the alligatoroid-crocodyloid clade found in morphology-only analyses, were not relevant). There is however a big problem with this.

Molecular and morphology based trees have long favored two competing phylogenetic hypotheses. When molecules are included or the only data used, gharial-like crocodylians (or gavialoids) are the sisters to Tomistoma, a gharial-like relative of crocodyloids (thus its common name, the false gharial), nesting them deep within the Crocodylia and putting alligator-like crocodylians on the outside. When morphology is the only data used, gharials sit at the base of the Crocodylia, with alligatoroids and crocodyloids being more closely related to each other.

Other authors have shown that in combined analyses, there is a high compatibility with the overall resolution of the crocodylian tree, ignoring the position of gavialoids. When molecular data is included, the internal resolution of Gavialoidea matches that found when morphology only analyses, and the tree is not otherwise disturbed. Since Salas-Gismondi et al. (2016) did not include molecular data or test their tree when molecular constraints are applied, I chose to do that for them. I used a molecular tree for Crocodylia[2–3] to constrain the relationships of the living crocodylians in the data set in PAUP* and then ran the data set with and without the constraint being enforced.

The molecular constraint only resulted in trees that were 1.6% longer than the morphology based ones and the tree was largely compatible with the unconstrained tree. In fact, in the constrained tree, the resolution of Gavialoidea was better, resolving the relationships of thoracosaurs, Eosuchus, and Eogavialis with respect to the remaining taxa. The only difference for non-gavialoid taxa was that all the species of Crocodylus represented formed a single clade (including two Eocene taxa, “C.” acer and “C.” affinis, which are thought to not represent Crocodylus proper [2-3]). It is possible this difference may reflect the poor sampling of crocodyloids in the matrix.

I think this small experiment suggests the need for caution when removing taxa from data sets. It would have been better to have had the additional new gavialoid related characters coded for the other crocodylian species and a continued move towards a bigger and better matrix of all crocodylians. However, this has no impact on the quality of the paper itself and is really just a pet peeve.

1. Salas-Gismondi, R., Flynn, J.J., Baby, P., Tejada-Lara, J.V., Claude, J., Antoine, P.-O. 2016. A new 13 million year old gavialoid crocodylian from proto-Amazonian mega-wetlands reveals parallel evolutionary trends in skull shape linked to longirostry. PLoS ONE, 11, e0152453. http://journals.plos.org/plosone/article?id=info%3Adoi%2F10.1371%2Fjournal.pone.0152453
2. Gatesy et al. 2003. Combined support for wholesale taxic atavism in gavialine crocodylians. Systematic Biology, 52: 402–423. http://sysbio.oxfordjournals.org/content/52/3/403.long
3. Brochu, C. A. 1997. Morphology, fossils, divergence timing, and the phylogenetic relationships of Gavialis. Systematic Biology, 46, 3, 479–522. http://sysbio.oxfordjournals.org/content/46/3/479.full.pdf