A note on the ability of STRUCTURE to correctly infer the number of populations for Bering-Chukchi-Beaufort Seas bowhead whales
Multiple stocks within the Bering-Chukchi-Beaufort region have been inferred from analyses of microsatellite data from bowhead whales killed on migration using the method STRUCTURE (Kitikado et al. 2007). We show that this conclusion is unwarranted and that STRUCTURE analyses are consistent with a s...
| Main Authors: | , , |
|---|---|
| Other Authors: | |
| Format: | Text |
| Language: | English |
| Subjects: | |
| Online Access: | http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.561.3485 http://www.iwcoffice.co.uk/_documents/sci_com/SC59docs/SC-59-BRG34.pdf |
| Summary: | Multiple stocks within the Bering-Chukchi-Beaufort region have been inferred from analyses of microsatellite data from bowhead whales killed on migration using the method STRUCTURE (Kitikado et al. 2007). We show that this conclusion is unwarranted and that STRUCTURE analyses are consistent with a single stock. We discuss model choice for STRUCTURE analyses and conclude that, based on bowhead biology, the appropriate model is one that assumes both recent ancestry and some current gene flow (admixture with correlated allele frequencies). Bowhead biology and history is not consistent with the model used by Kitikado et al. (2007) that assumes strong evolutionary separation and no current gene flow (no admixture and independent allele frequencies). We perform two new analyses to facilitate the appropriate selection of the number of populations (K) suggested by STRUCTURE analyses. The inference method currently recommended by STRUCTURE’s authors strongly suggests one stock within the BCB region. This is consistent with results using individual-based simulations of a single stock modeled to recreate BCB bowhead population dynamics and history (described in Archer et al. 2007): simulated datasets generated from a single stock and analyzed using STRUCTURE in the manner employed by Kitikado et al. falsely favored K=2 about 30 % of the time. We conclude that in the analysis of the full dataset, including BCB, Okhotsk and Atlantic samples, that STRUCTURE likely makes two errors: 1) it incorrectly fails to identify BCB and Atlantic whales as separate stocks, and 2) it incorrectly identifies two biologically meaningless groups within the pooled BCB-Atlantic samples if selection of K is done in a manner not supported by the current literature. |
|---|