Multibreed genomic prediction using summary statistics and a breed-origin-of-alleles approach

Area of Science:

• Animal Genetics
• Dairy Cattle Breeding
• Genomic Prediction

Background:

• Increasing interest in crossbreeding dairy cattle necessitates breeding values for crossbreds.
• Predicting genomic breeding values in crossbreds is complex due to differing genetic structures and potential data sharing limitations between breeds.
• Incomplete breed information can lead to low prediction accuracy for crossbred genetic merit.

Purpose of the Study:

• To investigate the accuracy of genomic predictions for crossbred dairy cattle using summary statistics from pure breeds.
• To evaluate a genomic prediction model incorporating breed-origin of alleles (BOA).
• To compare prediction accuracies using raw data versus summary statistics from reference populations.

Main Methods:

• Simulated two- and three-breed rotational crosses.
• Employed a genomic prediction model accounting for breed-origin of alleles (BOA).
• Compared prediction accuracies using full genotype/phenotype data versus summary statistics from pure breed genomic predictions.

Main Results:

• The BOA approach achieved prediction accuracies similar to a joint model when genomic correlations between breeds were high.
• Reference populations with summary statistics from all pure breeds and crossbred data yielded high prediction accuracies (0.720-0.768).
• Lacking pure breed information significantly reduced prediction accuracies (0.590-0.676).

Conclusions:

• Summary statistics from pure breeds can be effectively used for genomic predictions in crossbreds, particularly with a BOA model.
• Including crossbred animals in reference populations benefits purebred prediction accuracy, especially for smaller breeds.
• Access to comprehensive data from all contributing pure breeds is crucial for maximizing prediction accuracy in crossbred dairy cattle.