Determining the number of animals required to accurately determine breed composition using genomic data Proceedings

Using genomic data to assign breed or breed content is becoming commonplace in both the research setting and in genetic evaluation. There is potential to use this strategy for breed verification purposes where pedigree records are missing or incomplete. The current study provides guidelines on the number of cattle required in a reference to accurately assign breed to animals of unknown origin. A reference dataset comprising SNP genotypes (5817 loci) from Brahman, Hereford, Shorthorn and Angus were used to train the program Admixture and assign breed to individuals outside the reference set. The dataset demonstrates that as few as 50 animals are required to train the prediction to achieve accurate (>0.97) discrimination between breeds that are distantly related. However, at least 500 animals are required in the reference set to achieve accurate (>0.96) discrimination between closely related breeds such as Hereford and Shorthorn