Abstract #M62
Section: Breeding and Genetics (posters)
Session: Breeding and Genetics 1
Format: Poster
Day/Time: Monday 7:30 AM–9:30 AM
Location: Exhibit Hall A
Session: Breeding and Genetics 1
Format: Poster
Day/Time: Monday 7:30 AM–9:30 AM
Location: Exhibit Hall A
# M62
Impact of accounting for parent and genotyped daughters’ average in the estimation of deregressed estimated breeding values used in multiple-step genomic evaluations.
H. R. de Oliveira1,2, L. F. Brito2,3, M. Sargolzaei2,4, F. Fonseca e Silva1, J. Jamrozik2,5, D. A. L. Lourenco6, F. S. Schenkel*2, 1Universidade Federal de Viçosa, Viçosa, Minas Gerais, Brasil, 2University of Guelph, Guelph, ON, Canada, 3Purdue University, West Lafayette, IN, 4Select Sires Inc, Plain City, OH, 5Canadian Dairy Network, Guelph, ON, Canada, 6University of Georgia, Athens, GA.
Key Words: double-counting, genomic BLUP (GBLUP), training population
Impact of accounting for parent and genotyped daughters’ average in the estimation of deregressed estimated breeding values used in multiple-step genomic evaluations.
H. R. de Oliveira1,2, L. F. Brito2,3, M. Sargolzaei2,4, F. Fonseca e Silva1, J. Jamrozik2,5, D. A. L. Lourenco6, F. S. Schenkel*2, 1Universidade Federal de Viçosa, Viçosa, Minas Gerais, Brasil, 2University of Guelph, Guelph, ON, Canada, 3Purdue University, West Lafayette, IN, 4Select Sires Inc, Plain City, OH, 5Canadian Dairy Network, Guelph, ON, Canada, 6University of Georgia, Athens, GA.
Cows have been genotyped and used to increase the size of the training population in multiple-step genomic evaluations of dairy cattle with limited number of proven bulls. However, some of the genotyped cows are daughters of progeny tested bulls, which have their estimated breeding values (EBVs) predicted based on the parent average (PA) and the phenotypic information from their daughters, which would cause double count of information. The objective of this study was to investigate the impact of accounting for PA and genotyped daughters’ average (GDA; i.e., the contribution of genotyped daughters to the bull’s EBV) in the estimation of deregressed EBVs (dEBVs) used as pseudo-phenotypes in genomic evaluations. In addition, an alternative deregression method was proposed (NEW). A simulated dairy cattle data set was used to compare 8 scenarios defined based on the number of bulls, genotyped bull’s daughters, and genotyped cows not sired by the genotyped bulls. For all these scenarios, Genomic EBVs (GEBVs) were predicted using dEBVs estimated based on 4 methods: VR, that includes PA and GDA information in the dEBV; VRpa, that excludes PA; and JA and NEW, which exclude PA and GDA from the dEBVs using either all information available in the complete pedigree or only information from parents and genotyped daughters, respectively. The dEBVs estimated by the VR and NEW showed the lowest (0.24 to 0.36) and highest (0.33 to 0.50) validation reliabilities across scenarios, respectively. The VRpa and NEW methods produced the least biased GEBVs (inflation/deflation) and showed the most consistent bias estimates (regression coefficient) across scenarios (1.08 to 1.17). Among all methods, the JA method displayed the largest variability in bias (1.00 to 1.75) across scenarios. Therefore, it was shown that removing PA and GDA information from dEBVs can increase the reliability of genomic predictions for populations with limited number of proven bulls. In addition, the proposed NEW deregression method addresses the double counting of information and it is a feasible alternative to generate dEBVs used in multiple-step genomic evaluations.
Key Words: double-counting, genomic BLUP (GBLUP), training population
