2.3 Phylogenomic reconstructions
To explore species monophyly and patterns of micro-endemism using genome scale data (e.g., Wagner et al., 2013), evolutionary relationships amongNiebla specimens were inferred from RADseq data using two strategies, (i) a supermatrix approach from the concatenated RADseq loci and (ii) a computationally efficient species tree inference approach accounting for incomplete lineage sorting (Chifman and Kubatko, 2014) using variable sites extracted from RADseq loci. Concatenation, or phylogenomic supermatrix approaches, have been shown to accurately infer relationships across a range of scenarios (Tonini et al., 2015). We inferred relationships from both the concatenated RADseq loci, and a separate dataset comprised of only variable sites (SNP dataset). Phylogenetic trees were reconstructed using maximum likelihood as implemented in IQ-TREE v 1.6.12 (Nguyen et al., 2015), with 1,000 ultra‐fast bootstrap replicates (Hoang et al., 2018) to assess nodal support. For the SNP dataset comprised of 198K SNPs (see results), we used the GTR+ASC to avoid overestimated branch lengths due to the SNP alignment comprised only of variable sites. For the concatenated RADseq loci dataset, comprised of 25,086 loci, total alignment length 3.6 Mb (see results), we used the GTR+I+G substitution model. Because standard concatenation approaches may return incorrect trees with high support in the presence of incomplete lineage sorting (Edwards, 2009; Kubatko and Degnan, 2007), we used SVDQuartets, as implemented in PAUP*, a method that infers relationships among quartets of taxa under the multispecies coalescent model (Chifman and Kubatko, 2014). SVDQuartets+PAUP* is computationally efficient with large genome-scale datasets and able to accurately infer relationships under a range of scenarios (Chou et al., 2015). All possible quartets were evaluated under the multispecies coalescent tree model, and nodal support was performed using 100 bootstrap replicates.