Determining breeding origin of individuals on the nonbreeding
range
We assigned individuals sampled from the nonbreeding range to the ESSBPs
using WGSassign. Since these individuals are of unknown origin, we
assumed the accuracy of their assignment would be comparable to the
accuracy achieved with known breeding samples. Individuals at the
periphery or boundaries of genetically distinct populations may have
admixed genomes that are not truly representative of either population.
While posterior probabilities of assignment are typically used to detect
admixture and determine confidence of assignment, our preliminary
results showed that posterior probabilities of assignment were
unreliable with lcWGS data (see Results and Discussion ).
Therefore, we split up the genotype likelihood data into 10 subsets of
400,000 SNPs, in order of genomic region, for each individual and used a
consistency of assignment threshold of 0.8 (i.e. , at least 8 of
the 10 datasets being assigned to the same population) to determine
confidence in assignment. We validated this approach on the 47 breeding
individuals used as the testing set for the ESSBPs and then used it for
the 148 individuals from the nonbreeding range.