Genomic selection

Abstract

Computer simulation was used to examine the relationship between the number of loci included in the model and the accuracy of whole-genome methods for genetic evaluation. Random-regression BLUP (RR-BLUP) and Bayesian inference (Bayes-B) were compared with BLUP based on trait phenotypes and pedigree (TP-BLUP), where the genomic information consisted of genotypes at 110 or 2010 unlinked SNP loci. Of these loci, 10 were the actual trait loci. Accuracies for RR-BLUP were much higher than for TP-BLUP with 110 loci, but they were closer to TP-BLUP with 2010 loci. Accuracies with Bayes-B were close to 1 with 110 or with 2010 loci. Results for Bayes-B were also examined with a range of 3000 to 60 000 markers in linkage disequilibrium on 30 chromosomes. Increasing the number of markers did not lower the accuracy of Bayes-B. Thus, Bayes-B seems to be well suited for whole-genome analyses with large numbers of loci. This may not be the case for RR-BLUP.

Keywords:

• Bayesian inference
• BLUP
• computer simulation
• genetic evaluation
• marker-assisted selection

Acknowledgements

Support from the National Research Initiative Competitive Grants Program of the U.S. Department of Agriculture, Award 2007-35205-17862 (Fernando and Dekkers) and Newsham Genetics (Fernando) are gratefully acknowledged.