Beyond tree-like evolution in ostariophysan fishes

Phylogenetic trees have long been a powerful model of the evolutionary history. However, key processes in speciation, such as introgression and hybridization are difficult to depict with strictly binary trees alone, even with modern discordance metrics and post-hoc tests. Phylogenetic networks offer an intuitive way to estimate and interpret these reticulate events, as conflicting signal from gene trees should become evident as hybrid edges. Phylogenetic networks raise a number of challenges over tree-based analyses though, including the estimation of uncertainty and evaluating confidence among multiple competing network hypotheses. Here, we develop a new approach for assessing branch and topological support on networks using composite likelihood, and explore the utility of such statistics by revisiting competing evolutionary hypotheses in ostariophysan fishes. Our results suggest that a tree-like topology is not the best model for ostariophysans. A phylogenetic network with four reticulation events involving early-diverging lineages was the best alternative scenario. Our branch support metrics demonstrated non-negligible levels of uncertainty in the placement of those reticulation events though. An important question though, is if the uncertainty in the placement of these reticulation edges affects the qualitative conclusions of organismal evolution. To this end, we apply our network hypotheses to the biogeographic history of ostariophysans. Since no biogeographic models for networks exist, we implemented a parsimony approach using Fitch's algorithm for ancestral character state reconstruction. This represents the first biogeographic reconstruction for networks, and while simplistic, is a basis for further development. Biogeographic results are generally consistent with previous research on the presence of ancestral lineages in Laurasia and Gondwana post the break-up of Pangea, but also identify some cases of long-distance dispersal.