Using Phylogenetic Network Methods for Genomic Data Exploration and Hypothesis Generation Fails to Untangle a Confusing History of Hybridization in New Zealand Cicadas

Rapid species radiations make hybridization among species more likely. Detecting and reconstructing hybridization is therefore critical for understanding species relationships in many cases. We explored the relative performance of two phylogenetic network methods, SNaQ, a gene tree-based method, and PhyNEST, a site pattern-based method, in evaluating the plausibility of proposed past hybridization hypotheses. As our study system, we used the New Zealand cicada genera Kikihia and Maoricicada. Previous phylogenomic work on these two species radiations suggested multiple hybridization events in response to changing landscapes and climate. We generated hypotheses for specific hybridization events based on observed hybrid mating songs and patterns of mito-nuclear discordance from previous studies. We tested our hypotheses using the D-statistic and a phylogenomic dataset of over 500 nuclear Anchored Hybrid Enrichment genes along with mitochondrial genomes. This larger dataset provided stronger support for some of our hybridization scenarios but not all. Using these same data we inferred phylogenetic networks using SNaQ and PhyNEST to determine whether the two methods recovered plausible network with respect to our hypothesized hybridization events. We found that both SNaQ and PhyNEST recovered an extensive history of reticulate evolution in New Zealand cicadas which broadly matched our predictions. We suggest that differences between networks inferred by the two network programs may result from using site patterns versus gene trees as input data or reflect other differences in the inference methods. Finally, we discuss considerations for users applying these methods to targeted enrichment data and suggest improvements for network method developers.