Genetic diversity and population structure
We estimated diversity statistics within populations using ANGSD (Korneliussen et al., 2013). First, we calculated nucleotide diversity as the average number of pairwise differences (π) (Nei & Li, 1979) and as the proportion of segregating sites (θW) (Watterson, 1975)⁠. We calculated heterozygosity for each individual based on its site frequency spectrum (SFS) and estimated the inbreeding coefficients (F) using ngsF (Vieira et al., 2013)⁠.
To assess population structure using PCA, we created a covariance matrix among individuals using ngsCovar from the ngsTools suite (Fumagalli et al., 2014)⁠ and calculated principal components in R v3.4.4 (R Core Team, 2018)⁠ using the ‘eigen’ function. The number of principal components explaining most of the population structure was determined from the scree plot of PCA (Cattell, 1966)⁠.
We assessed admixture among populations using NGSadmix (Skotte et al., 2013)⁠ with a number of clusters K ranging from 2 to 14. We repeated each analysis 20 times and reported the results of the highest likelihood analysis for each K . Finally, we calculated the pairwise FST in ANGSD using the shared site frequency spectrum for each pair of populations. The results were also used to test for isolation by distance (IBD) by calculating the correlation between pairwise linearized FST values [FST/(1-FST)] and log-transformed pairwise geographic distance (Rousset, 1997) using a Mantel test with 1,000 permutations in the R package ’Vegan 2.5-4 ’ (Dixon, 2003; Table S4)