Implementation of GPS-MACE accounts for genomic preselection


  • Peter G Sullivan Canadian Dairy Network
  • Esa Mäntysaari Natural Resources Institute (Luke)
  • Gerben de Jong CRV u.a.


Genomic pre-selection (GPS) has altered the distributions of breeding values for AI bulls, because genomics made it possible to identify above average bulls within a family prior to progeny testing.  Before genomics, it was reasonable to assume within-family pre-selection was random in EBV models, but this assumption is no longer valid.  The international MACE model was thus modified to account for non-random within-family GPS of AI bulls.  The effects of GPS are estimated and included in the international EBV of sires in the new model: GPS-MACE.  The estimates reflect different intensities of GPS across traits, breeds, countries of selection and time, and the accumulation of differential effects of GPS across multiple generations in a bull’s ancestry, and across international borders.  Estimates of GPS effects, and differences between EBV from GPS-MACE versus MACE were studied for three breeds (Holstein, Brown Swiss, and Jersey) and eight traits (milk fat, milk protein, conformation score, udder score, somatic cell score, fertility interval, cow conception, and milking speed).  The effects of GPS were generally largest for countries that have shared genotypes of proven bulls to improve national genomic predictions and the effectiveness of national GPS programs.  The countries identified as sharing genotypes were in the Inter-continental and Eurogenomics Holstein consortia, and the Intergenomics Brown Swiss service provided by Interbull.  Estimates of GPS effects were largest and almost always favourable across all breeds and countries for the main traits of selection (fat, protein, conformation score, and udder score), and were generally centred on zero for the traits under weaker selection intensities (milking speed).  The addition of GPS effects to the international model caused generally higher estimates of Mendelian sampling effects and correspondingly lower EBV for the dams of GPS bulls, when the estimated effects of GPS were positive.  The net effects on EBV were small for older GPS bulls with high national EBV reliabilities, but they were notably larger and favourable for the most recent AI bulls from countries with effective GPS programs.  The benefits of GPS-MACE over MACE were likely underestimated in the present study because GPS effects are only partially included in the current national EBV.  As national EBV models are improved to more fully account for the effects of GPS, benefits of using GPS-MACE are expected to grow.