PhD: Seeing the trees for the forest: a comparison of methods for inferring the tree of life

A central goal of biology is to uncover the tree of life – determining how all species, extinct and extant, are ultimately related. Here the student will investigate how different approaches to inferring this tree can be compared. Pre-existing databases of primarily animal trees, but with strong focuses on arthropods (e.g., insects, crustaceans) and tetrapods (e.g., birds, dinosaurs, ichthyosaurs, lizards, plesiosaurs, pterosaurs, mammals) will be expanded and developed, allowing the broadest possible set of comparisons to be made. The student will also have the option to develop tree(s) of their own. Comparisons will be made using methods that compare the branching order of trees with the appearances of taxa in the fossil record. Thus only groups with fossil records will be used.

Because the branching order of different phylogenies and the appearance of taxa in the fossil record are independent estimates of the same thing their mutual agreement is suggestive of accuracy (and by extension, disagreement of inaccuracy). A suite of metrics has been produced to make such comparisons (summarised in Bell and Lloyd 2015). These allow competing phylogenies to be compared. Although so far limited headway has been made in this area (but see Brochu and Norell 2001). Instead these metrics have been applied to different groups and time periods to see if any major patterns emerge (for a recent summary see O’Connor and Wills 2016).

Comparison of competing phylogenies is important as there is disagreement on the best methods to use in inferring them. Palaeontologists have traditionally relied on parsimony – algorithms that attempt to select trees that minimise the number of evolutionary “steps”. However, more recently model-based approaches that make particular assumptions about evolutionary change have been employed. Similarly, different statistical paradigms are in use. For example, the Bayesian approach, which requires a number of guesses to be made about the tree before analysis proceeds. Yet other approaches combine the outputs of individual analyses into larger “supertrees” (Davis and Page 2014; Hill and Davis 2014; Lloyd et al in press). Being able to compare the outputs of trees generated from such disparate approaches should help aid in establishing a “best practice” in the field, and hence the project has the potential to be highly influential.

The student will primarily use the software package “strap” (Bell and Lloyd 2015) to make their comparisons, for which the main supervisor is an author. Other software produced by, or otherwise familiar to the supervisors will also be employed (e.g., The Supertree Toolkit 2; Hill and Davis 2014), allowing the student the best possible access to cutting edge techniques.