Modern individual clustering methods utilising
hypervariable nuclear
microsatellite DNA polymorphisms are being increasingly applied in the field of
population genetics. This study explores the efficiency of the clustering methods in identifying the
breeds of origin of 250 domestic dog (
Canis familiaris) individuals based on 10
microsatellite loci. An
allele sharing distance (DAS) matrix and the corresponding
neighbour-joining tree of individuals revealed
monophyletic assemblages that corresponded perfectly with the
breeds of origin of the dogs. Individual assignment tests using a
Bayesian statistical approach, an
allele frequency based method, and a DCE
genetic distance based method were all extremely powerful. Most strikingly, the
Bayesian method provided 100% assignment success of individuals into their correct
breeds of origin and 100% exclusion success of individuals from all alternate reference populations with a high level of
statistical confidence (P < 0.0001). A
Bayesian Markov Chain Monte Carlo clustering approach revealed clear distinction of individuals into groups according to their
breeds of origin, with a near-zero level of '
genetic admixture' among
breeds. The results demonstrate that an FST of 0.18,
mean expected gene diversity of 0.6 across 10 loci, and approximately 50 individuals per reference population suffice to provide maximum individual assignment success in C. familiaris. This refutes the traditional view that
DNA based
dog breed identification is not feasible at the individual level of resolution.
