Dynamic Genomic Selection in Crossbred Beef Cattle Populations

ABSTRACT:

Accuracy of genomic selection has been shown to be proportional to the number of animals in the genotyped training population as well as the relationship between training and prediction populations. Adding extra training animals of breeds that are not closely related to the animals in the prediction population may lead to further inaccuracy in estimation of SNP effects, as SNP effects may not be consistent across populations. This study proposes a novel, unbiased method of creating clusters of animals to be evaluated together based on the Euclidean Distance Matrix (EDM). This clustering method was tested on real crossbred beef data, and compared to evaluations incorporating all available animals into one common population. Increased accuracy (>0.057) was obtained when the training population was defined based on the clustering methodology. These preliminary results warrant further investigation.

Keywords:
beef cattle

accuracy

clustering