Alternative Approaches to handling of missing parents in genetic evaluation of dairy cattle using single-step test-day SNP-BLUP model

Abstract

In many countries, single-step genomic models are replacing conventional pedigree-based models for routine valuation. Those models use all available information on the animals’ phenotype, genotype, and pedigree. Pedigree data still has a huge impact on estimated genomic breeding values (GEBV), and it is also important to consider information about the structure of the pedigree. The foremost aspect of pedigree editing is dealing with missing parents' information. The choice of method of handling missing parents can affect the prediction of breeding values. This work investigates three scenarios of pedigree data: 1) Pedigree_real (P_Real) – pedigree from the routine evaluation, 2) Pedigree_2010 (P_2010) – at least 20 and 10 percent of dams and sires born before 2019 were set randomly to missing, respectively, 3) Pedigree_4020 (P_4020) – at least 40 and 20 percent of dams and sires born before 2019 were set randomly to missing, respectively. Moreover, for those pedigrees, three approaches to defining missing parents were used:  1) Raw pedigree (RP) – missing parents IDs set to missing, 2) Genetic groups (GG) – missing parents replaced by unrelated GG, which are defined based on year of birth, sex, and country of origin, 3) Metafounders (MF) – missing parents replaced by MF, which correspond to genetic groups. Relationships within and between metafounders were estimated from genomic information of descendants. The genomic breeding values for fat yield were estimated using the single-step test-day SNP-BLUP model, implemented by the MiXBLUP software. Although GEBV prediction was similar across scenarios, expressing missing parents by GG or MF impacts the genetic trend, especially in situations of limited pedigree completeness. Removing parent information led to reduced precision results across the methods of handling missing parents, since P_Real scenario demonstrated highest accuracy results. Compared to RP and GG, MF scenarios resulted in higher genetic trends. Insufficient pedigree completeness, especially among ungenotyped individuals, leads to an overestimation of the genetic trend. Completeness of pedigree information and a large number of genotyped individuals improve the reliability of evaluations. Modeling missing sires with MFs is less effective than assuming unrelated GGs if pedigree information is very incomplete. Therefore, the best method to model missing parents depends on completeness of pedigree.