Usefulness of the clustering methodologies to discriminate between purebred and crossbred individuals

Abstract

Molecular markers have been successfully used to distinguish between livestock species and breeds not closely related, for example through the clustering methodology. However, the differentiation between purebred and crossbred individuals would be an appealing purpose that has been little explored. In this study three clustering approaches are tested for their ability to detect crossbred individuals and to separate them from pure ones. Real microsatellite data from Iberian and Duroc breeds were utilised as an example. Simulated F1, Iberian and Duroc backcrossed individuals obtained from the real microsatellite were also assessed. The results of this study indicate that the clustering methods showed a reduced ability to detect the original subpopulations (Iberian breed, Duroc breed, F1, Iberian backcross and Duroc backcross). Reasons for such performance could be the absence of Hardy-Weinberg and linkage equilibrium within the subpopulations and the fact that the Iberian group was compound by individuals belonging to different strains. To test the influence of these factors an allele randomisation procedure was performed within each subpopulation. After that, none of the methods recovered the five groups, but the algorithm implemented in BAPS (Bayesian analysis of population structure) gave a partition where pure Iberian individuals were separated for the rest. It can be concluded that the lack of homogeneity within groups is the main cause of the reduced accuracy of the clustering methods in the separation of pure and crossed individuals.