Development of a projection-based method for genomic selection in admixed and crossbred populations

Summary

A projection based approach was developed for dealing with admixed and crossbred populations in genomic selection.

Situation

Currently, genome wide association studies and genomic selection are often conducted using purebred populations. Estimation and often validation of SNP effects are carried out using a select elite set of purebred animals (i.e. proven sires). This process is successful when estimated SNP effects are used to predict genomic breeding values (or pseudo-phenotypes) on animals of the same breed or very closely related animals. However, it fails at various degrees when these SNP estimates are used for genomic prediction in different breeds or crossbred animals. This situation is very challenging to segments of livestock industry (beef cattle, swine or poultry) where the traits of interest are measured in crossbred or mixed population with uncertain breed composition. Current approaches for dealing with admixed and crossbred populations in genomic selection rely on using different groups of animals in the training and validation sets, and hence they are data dependent and often lead to reduction in accuracies for animals in the pure breed populations.

Response

Our idea is to use SNP effects estimated using purebred animals to infer their estimates in crossbred populations. It is obvious that changes in SNP effects between populations and crossbred is due to several factors including, change of minor allele frequency, and strength of LD between markers and QTLs, and linkage phase between marker and QTL alleles. Our hypothesis is that, given a breed composition of a given individual, is it possible to estimate the SNP effect in this crossbred animal through projection. Pure breed, phenotypes (or pseudo-phenotypes) were normalized within each breed. This step is needed to remove differences in SNP estimates between breeds that are due solely to the magnitude and dispersion of the trait values. Using the normalized trait values, SNP effects were estimated for each breed separately. Having the SNP effects in the purebreds and the composition of the crossbred animals, the estimates of an SNP effect in subpopulation of crossbred animals is obtained indirectly through a projection equation. To evaluate our procedure, a simulation involving two breeds and their crossbreds (50/50, 25/75, and 75/25%) was conducted.

Impact

Assuming the knowledge of the breed composition of each crossbred individual, our procedure was used to first estimate the SNP effects separately in each of the two pure breeds. Using our projection equation, SNP effects in each of the three crossbred population were obtained. Using these estimates, the pseudo-phenotypes were calculated. The accuracies, measured as the correlation between true and estimated pseudo-phenotypes, for the three cross breed population ranged between 0.29 and 0.43. These correlations were more than two-fold higher compared to those obtained using the purebred SNPs estimates directly. Furthermore, they were 20 to 51% higher compared to using different mixture of purebred and crossbred animals in training and validation steps as currently done. Although the results are promising based on simulated data, our procedure is being evaluated using real chicken data.

State Issue

Agricultural Profitability and Sustainability

Details

• Year: 2011
• Geographic Scope: International
• County: Clarke
• Program Areas:
  • Agriculture & Natural Resources