Ceridwen I. Fraser, Angela McGaughran, Aaron Chuah, Jonathan M. Waters

Genome-wide SNP data generated by non-targeted methods such as RAD and GBS are increasingly being used in phylogenetic and phylogeographic analyses. When these methods are used in the absence of a reference genome, however, little is known about the locations and evolution of the SNPs. In using such data to address phylogenetic questions, researchers risk drawing false conclusions, particularly if a representative number of SNPs is not obtained. Here, we empirically test the robustness of phylogenetic inference based on SNP data for closely-related lineages. We conducted a genome-wide analysis of 75,712 SNPs, generated via GBS, of southern bull-kelp (Durvillaea). Durvillaea chathamensis co-occurs with D. antarctica on Chatham Island, but the two species have previously been found to be so genetically similar that the status of the former has been questioned. Our results show that D. chathamensis, which differs from D. antarctica ecologically as well as morphologically, is indeed a reproductively isolated species. Furthermore, our replicated analyses show that D. chathamensis cannot be reliably distinguished phylogenetically from closely-related D. antarctica using subsets (ranging in size from 400 to 40,912 sites) of the parsimony-informative SNPs in our dataset, and that bootstrap values alone can give misleading impressions of the strength of phylogenetic inferences. These results highlight the importance of independently replicating SNP analyses to verify that phylogenetic inferences based on non-targeted SNP data are robust. Our study also demonstrates that modern genomic approaches can be used to identify cases of recent or incipient speciation that traditional approaches (e.g., Sanger sequencing of a few loci) may be unable to detect or resolve. This article is protected by copyright. All rights reserved.