Revision of random regression test-day model has improved genomic prediction for Nordic Red dairy cattle

Abstract

In 2006, Denmark, Finland, and Sweden have introduced across-country genetic evaluations for yield traits for their Holstein, Red dairy cattle (RCD) and Jersey populations. The implemented breed-specific random regression test-day models (RRM) were the outcome of an intensive research cooperation among the countries. Especially developing the RRM for the RDC population presented unique challenges due to its heterogeneous population structure spanning across Finland, Sweden, and Denmark. As genomic prediction became a key tool for breeding decisions, it became evident that the reliability of genomic enhanced breeding values (GEBV) for the RDC breed was lower than expected when compared to the Holstein and Jersey breeds. Several factors contributed to this discrepancy. Notably, during the last two decades, changes in herd and population structures were most pronounced within the RDC breed where the original RDC country-subpopulations have become much more alike. Thus, revision of the RRM is crucial in enhancing the reliability of GEBV for the selection of breeding candidate animals. First important improvements were the revision of modelling automated milking system data and a newly estimated set of variance components with lower h². The updates made so far indicate considerable improvement in the genomic predictions. The LR regression coefficient (b₁) values increased for example for milk yield from 0.85 of the original model to 0.92 of the revised model, indicating that the bias decreased with the revised model. Also, the coefficient of correlation (R²) increased for all production traits on average 4.5%. In a next step, we will truncate the phenotypic data, optimize the pedigree information, and study whether modelling metafounders for the heterogeneous RDC population will result in further improvements for the genomic prediction.