Population genetic analyses
To assess phylogeographic structure, a genetic distance matrix was computed using ngsDist using genotype likelihoods (Vieira et al. , 2016). Following recommendations of RAxML (Stamatakis, 2014), a bootstrapped neighbor-joining tree matrix was computed from 1,000 possible trees and visualized as a phylogenetic tree using FASTME (Lefort, Desper and Gascuel, 2015) and FigTree v.1.4.2 (Rambaut, 2009).
We estimated the within-population genetic diversity of Lineage B using two diversity measures, namely expected heterozygosity and nucleotide diversity. We calculated expected heterozygosity (He) using GENODIVE (Meirmans and Van Tienderen, 2004), and overall heterozygosity and nucleotide diversity (π) using ANGSD (Nei, 1987).
Population structure was identified using three methods. First, we performed a neighbor-joining network (NeighborNet) analysis using Splitstree (Huson, 1998; Huson and Bryant, 2006). Splitstree does not force a tree-like structure onto the data and thus can verify the extent to which the data conform to a hierarchical tree structure. Next, we ran a Principal Component Analysis (PCA) based on a covariance matrix computed by ngsTools on genotype likelihoods (Fumagalli, Vieira and Linderoth, 2014) and via GENODIVE using genotype calls. As an unsupervised clustering method, PCA estimates population genetic structure in an unbiased way. Finally, we explored admixture patterns using ngsAdmix (Skotte, Korneliussen, 2013). Ancestry of populations was explored through calculating admixture proportions per individual and varying the estimated number of ancestral populations (K). The most likely K was determined by running 10 replicate runs of each respective K, calculating the log likelihood value of each, and choosing the value of K where an addition of an ancestral group did not result in a higher likelihood (Evanno, Regnaut and Goudet, 2005).
Normalized population differentiation was calculated using high confidence genotype calls in GENODIVE. Normalized fixation index (F’ST) was calculated to eliminate the effect of within-population diversity (Meirmans and Hedrick, 2011). Finally, we constructed a migration network using Nei’s GST with the threshold at 0.4 using the DiveRsity package in R (Keenan et al. , 2013), as demonstrated by Sundqvist et al. (2016).