Optimum multistage genomic selection in dairy cattle

Abstract

The availability of different single nucleotide polymorphism (SNP) chips and the development of imputation algorithms allow for multistage dairy cattle breeding schemes applying various genomic selection strategies. These SNP genotypes yield genomic ally estimated breeding values (GEBV) with different accuracies at different costs. Thus, the optimum allocation of investments to different selection paths and strategies to maximize the genetic gain per year (ΔGa) and its sensitivity to changes in cost and accuracies of GEBV is of great interest. This is even more relevant under the constraints of limited financial resources. With deterministic methods, optimum multistage breeding plans maximizing ΔGa were identified in which selection could take place on GEBV derived from high-density (GEBVHD) and low-density (GEBVLD) SNP genotypes. To account for the uncertainty of cost and accuracies of GEBV, these parameters were varied in a semi-continuous manner. Overall breeding costs were limited to the crucial expenses of a traditional breeding program with 50 progeny-tested young bulls per year. Results clearly show that, in an optimal selection strategy, selection on GEBVLD is predominantly used for the identification of future bull dams but the main part of ΔGa is still generated from selection of sires. The low selection intensity in the path dam to sire induced a higher sensitivity of ΔGa to changes in cost and accuracies of GEBVLD compared with the same changes of GEBVHD. On the contrary, the genetic gain generated from selection of males was only affected by changes in accuracies of GEBVHD but almost unaffected by any changes in cost. Thus, changes in cost and accuracies of GEBVLD put the most pressure on the breeding scheme structure to maintain a high ΔGa. Furthermore, genomic selection of bull dams produced by far the majority of breeding cost but the lowest genetic gain.