Evaluating methods for phylogenomic analyses, and a new phylogeny for a major frog clade (Hyloidea) based on 2214 loci.

Phylogenomic approaches offer a wealth of data, but a bewildering diversity of methodological choices. These choices can strongly affect the resulting topologies. Here, we explore two controversial approaches (binning genes into "supergenes" and inclusion of only rapidly evolving sites), using new data from hyloid frogs. Hyloid frogs encompass ∼53% of frog species, including true toads (Bufonidae), glassfrogs (Centrolenidae), poison frogs (Dendrobatidae), and treefrogs (Hylidae). Many hyloid families are well-established, but relationships among these families have remained difficult to resolve. We generated a dataset of ultraconserved elements (UCEs) for 50 ingroup species, including 18 of 19 hyloid families and up to 2214 loci spanning >800,000 aligned base pairs. We evaluated these two general approaches (binning, rapid sites only) based primarily on their ability to recover and strongly support well-established clades. Data were analyzed using concatenated likelihood and coalescent species-tree methods (NJst, ASTRAL). Binning strongly affected inferred relationships, whereas use of only rapidly evolving sites did not (indicating ∼87% of the data contributed little information). The optimal approaches for maximizing recovery and support of well-established clades were concatenated likelihood analysis and the use of a limited number of naive bins (statistical binning gave more problematic results). These two optimal approaches converged on similar relationships among hyloid families, and resolved them with generally strong support. The relationships found were very different from most previous estimates of hyloid phylogeny, and a new classification is proposed. The new phylogeny also suggests an intriguing biogeographical scenario, in which hyloids originated in southern South America before radiating throughout the world.