Genomic Predictions Using Whole Genome Sequence Data and Multi-breed Reference Populations

Abstract Text: The availability of whole-genome sequence data (WGS data) on large number of livestock’s provides new opportunity for genomic selection. We investigated how much accuracy is gained by using WGS data in diverged cattle populations, using simulation. Relative performance of genomic BLUP and a Bayesian (BayesP) method with a mixture of normal distributions were compared. WGS data increased accuracy (3-7%) of within population predictions for moderate – lowly heritable traits. The advantage of WGS data (18-24%) was more pronounced with reference populations (RP) combined across breeds and when using BayesP. Extending the RP to multiple-breeds resulted in a 10-22% increase in accuracy with WGS data. BayesP outperformed GBLUP at 45 QTL/M, although in real data both methods have been shown to perform quite similar. Genomic predictions in numerically minor cattle populations would benefit from a combination of WGS data, multi-breed RP, and Bayesian estimation methods.

Keywords: genomic prediction; whole-genome sequence; multi-breed