2.3 Assessing genetic structure
We used multivariate, Bayesian, and admixture-based analyses to assess
population structure. In all analyses, clustering algorithms were run on
three data sets separately for comparison (T. blandingii ,T. pulverulenta , and both species combined [genusToxicodryas ]). A discriminant analysis of principal components
(DAPC) was run using Adegenet v. 2.1.1 (Jombart & Ahmed, 2011). This
approach uses discriminant functions to maximize variation among
clusters and minimize variation within clusters. The best-clustering
scheme was chosen based on Bayesian information criterion (BIC) scores.
Numbers of clusters (K) ranging from 1–10 were evaluated and a
discriminant function analysis of principal components (DAPC) was
performed based on the number of suggested clusters. Ancestry
proportions of all individuals were inferred using LEA v. 1.6.0 (Frichot
& François, 2015) through the Bioconductor v. 3.4 package. The sNMF
function was used to assess K values from 1–10, with 20 replicates,
estimate individual admixture coefficients, and select the value of
K that minimized cross entropy (François, 2016; Frichot, Mathieu,
Trouillon, Bouchard, & François, 2014). Population structure and
admixture were also tested using the Bayesian method STRUCTURE v. 2.3.4
(Falush, Stephens, & Pritchard, 2003; Pritchard, Stephens, & Donnelly,
2000). Each data set was evaluated for K=1–10 with 10 runs per K and a
MCMC burn-in of 10,000 steps followed by 100,000 steps (Porras-Hurtado
et al., 2013). Results were evaluated using the Evanno method (Evanno,
2005) and plotted through the R package pophelper v. 2.3.0 (Francis,
2017).