3.1 Population divergence and diversity using SNP array data
A total of 77 of the 96 SNPs were polymorphic in Lakes Bunnersjöarna. The average call rate of these loci was 0.989 (range: 0.707-1.0). The average per-individual-call-rate for these loci was 0.989 (range: 0.935-1.0). When grouping the individuals into demes based on theLDH-1 genotype, we found a strong genetic divergence between the populations; Weir & Cockerham´s (1984)F ST=0.24 (CI: 0.19-0.29), and Nei´s (1973) F ST=0.12 (CI: 0.10-0.14). The correlation between these two approaches for measuring F ST is high (r =0.99, p< 0.001). We found pronounced differences in the amount of genetic variation within the demes, consistent with data from the eight polymorphic allozyme loci (Ryman et al., 1979); only 24 of the 77 SNPs loci were variable in Deme II and expected heterozygosity is 0.27 for Deme I and 0.08 for Deme II (Table 1).
When analyzing the SNP array data without prior grouping of individuals, STRUCTURE suggested K =2 clusters, and these two clusters are almost completely consistent with the LDH-1 groupings (Figure 2a). Only one individual of Deme I (from LDH-1 genotype) was assigned to the other deme, but the assignment probability was very close to 0.5 (Q = 0.502). Kfinder suggested K =3 as the most likely number of clusters, but the STRUCTURE barplot for K =3 did not indicate a third separate cluster (Figure 2b). Further, the log likelihood approach from STRUCTURE suggests K =11, where four of these clusters occur primarily within allozyme LDH-1 Deme II and the remaining seven in the LDH-1 Deme I (Figure 2c, also see Figure S3). BAPS suggested a total of six clusters and five of these clusters were located within LDH-1 Deme I whereas the sixth cluster coincided completely with LDH-1 Deme II (Appendix S2).
An individual-based neighbor-joining tree including 127 individuals with full genotypes in 70 of the 77 polymorphic SNP loci further illustrated the clear partition of the two demes and showed additional substructuring within Deme I (Figure 3). From these observations, we conclude that the two demes identified by allozyme LDH-1 are supported by SNP array data.