Strategy to stabilize genomic breeding values under an evolving sire and cow reference population in the single-step evaluation system of Walloon region of Belgium

Abstract

The national genomic evaluations for production, conformation, udder health and functional traits in Wallonia are official since 2015. Nearly all evaluated traits are submitted to Interbull three times a year which give gaps of four months between official genomic estimated breeding values (GEBV). Generating reliable GEBV is a major challenge. With our small population, changes in the reference population size can be extremely important. Currently, approximately 12 000 Holstein genotypes are available, with 8 500 being actually used in the evaluations. However, through projects and intensive testing an increase of at least 20% per year is expected. The question is now how can we stabilize GEBV when the reference population is constantly moving? This research here is associated to the derivation of an interim computational method intended to help breeders to make early decisions. This implementation consisted in: GEBV partition into polygenic (PT) and direct genomic (DGV) values; SNP effect estimation from DGV; GEBV prediction for new animals by combining DGV generated from SNP effects and PT. We also investigated the hypothesis that a core group of animals would be sufficient to estimate GEBV for other non-reference animals. To test this, a list of 648 genotyped animals having official GEBV from the last run was used as validation. Interim GEBVs were generated (by summing up PT, DGV and mean trait), and correlated with their official GEBV. Correlations between official and interim GEBVs were 0.92, 0.93, 0.93, 0.93, 0.94, 0.91, 0.91, 0.94 and 0.94 for milk yield, fat yield, fat percentage, protein yield, protein percentage, somatic cell score, longevity, direct calving ease and maternal calving ease, respectively. Relative mean differences were up to 6%. These results are a first indication that we could develop a stable reference population, generate high quality SNP effects and generate appropriate GEBV reflecting potentially own records for non-reference population animals. The last point is joint with efforts to generate appropriate reliabilities based on the approach promoted by Interbull.