Resolving patterns of ancient and rapid cladogenesis is one of the hardest challenges in evolutionary biology, mostly due to difficulties associated with incomplete lineage sorting and homoplasy. Phylogenomic analysis of hundreds to thousands of loci offers promise to resolve these contentious relationships, but whether all the data that is generated contain useful phylogenetic information is uncertain. Here we assess the utility of different data filtering approaches to maximize phylogenetic information and minimize noise for resolving an ancient radiation of Neotropical electric knifefishes of the order Gymnotiformes, using ultraconserved elements. We recovered two contrasting evolutionary hypotheses depending on whether phylogenetic inference was based on concatenation or coalescent methods. In the first case, all the analyses recovered a previously—and commonly—proposed hypothesis, where the family Apteronotidae was the sister group to all other gymnotiform families. Coalescent-based analyses recovered a novel hypothesis where families producing pulse-type (Gymnotidae, Hypopomodiae and Rhamphichthyidae) and wave-type electric signals (Apteronotidae, Sternopygidae) were reciprocally monophyletic. Nodal support for this hypothesis increased when analyzing loci with highest phylogenetic information content, and most significantly when data were pruned using targeted filtering methods that maximized phylogenetic informativeness at the deepest nodes of the Gymnotiformes. Bayesian concordance analyses, and topology tests of individual gene genealogies, evidenced that the difficulty of resolving this radiation was owed to high gene tree incongruences most likely due to incomplete lineage sorting. Furthermore, we proved that data filtering reduces gene tree heterogeneity, and increases nodal support and consistency of species trees using coalescent methods, but not in concatenation.
Less...