Skip to main content
ARS Home » Research » Publications at this Location » Publication #223524
Title: Strategies to incorporate genomic prediction into population-wide genetic evaluations

Author
item GENGLER, N - GEMBLOUX AGRIC UNIV
item Vanraden, Paul

Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 2/28/2008
Publication Date: 7/11/2008
Citation: Gengler, N., Van Raden, P.M. 2008. Strategies to incorporate genomic prediction into population-wide genetic evaluations. Journal of Dairy Science. 91(E-Suppl. 1):506(abstr. 523).

Interpretive Summary:

Technical Abstract: Most current research on genomic selection is focusing on the accurate prediction of genomic breeding values. However selection solely based on genomic breeding values, despite being theoretically promising, is in practice only suboptimal for several reasons. The two most important are that only few animals are genotyped therefore having genomic prediction directly available and that rankings will change. With genomic breeding values potentially available in the near future, strategies are required to avoid any confusion in the mind of users. The aim of this study is to present three different strategies that could be used to incorporate genomic prediction into population-wide genetic evaluation. The three strategies are: 1) using selection index theory to combine both sources of information into a single set of breeding values; 2) for ungenotyped animals, compute conditional expectation of gene contents for SNP given molecular and pedigree data and use these predicted gene contents; and 3) integrate genomic breeding values as external information into genetic evaluation using a Bayesian framework. If strategy 1) is straight forward, additional steps have to be done to adjust breeding values for changes in those of relatives. A practical implementation is to use reliabilities of the genomic prediction, the population-wide genetic evaluation PA, and PA from the genotyped subset to set up a 3 x 3 matrix for each animal, with off-diagonal elements being functions of the 3 reliabilities. The use of strategy 2) is computationally much more challenging but leads directly to the needed covariance structures combining genomic relationship if known with pedigree relationships. Strategy 3) is potentially a good compromise because the theory is well established and has already been used in beef cattle to incorporate external breeding values. Also current genetic evaluation software can be easily modified to incorporate genomic breeding values.