SNP analyses
Genetic similarity among individuals and populations was visualized using ordination (Principal Coordinates Analysis [PCoA]; Gower, 1966), using individuals as entities and loci as attributes and implemented by the gl.pcoa and gl.pcoa.plot functions of dartR. For our phylogenetic analysis we constructed SNP genotypes for each individual by concatenating only the variable bases from each SNP locus into a single partition. A few loci had the SNP removed with the adaptor, because of chance matching of the adaptor sequence to the terminal region containing the SNP. These loci were removed prior to concatenation. Heterozygous SNP positions were represented by the standard ambiguity codes. We generated a phylogenetic tree using Maximum likelihood (ML) applied to concatenated sequences. ML analyses were conducted using RAxML 8.2.12 (Stamatakis 2014) on the CIPRES cluster (Miller, Pfeiffer & Schwartz, 2010) using the model GTRCAT and searching for the best-scoring ML tree using the model GTRGAMMA in a single program run, with bootstrapping set to finish based on the autoMRE majority rule criterion. The tree was imported to Mega 7.0.18 (Kumar, Stecher & Tamura, 2016), formatted and mid-point rooted. To assist with identifying potential introgressed individuals, heterozygosity was calculated in R using the command “het <- rowMeans(as.matrix(gl)==1, na.rm=T)” followed by “write.csv (het, file=”het.csv”)”.
The diagnosability of lineages and candidate species was assessed by calculating the number of pairwise fixed differences (both absolute and allowing a 5% tolerance for shared alleles at each locus) and the associated probabilities that such values could arise through sampling error alone (dartR command gl.fixed.diff; parameter tloc = 0 or tloc = 0.05; see Unmack et al., 2022 for rationale and methods involved).