|CESARANI, ALBERTO - University Of Georgia|
|LOURENCO, DANIELA - University Of Georgia|
|MASUDA, YUTAKA - University Of Georgia|
|LEGARRA, ANDRES - Inrae|
|TSURUTA, SHOGO - University Of Georgia|
|NICOLAZZI, EZEQUIEL - Council On Dairy Cattle Breeding|
|MISZTAL, IGNACY - University Of Georgia|
Submitted to: Journal of Dairy Science
Publication Type: Abstract Only
Publication Acceptance Date: 4/8/2021
Publication Date: 6/28/2021
Citation: Cesarani, A., Lourenco, D., Masuda, Y., Legarra, A., Tsuruta, S., Nicolazzi, E., Van Raden, P.M., Misztal, I. 2021. Multi-breed genomic evaluation for dairy cattle in the US using single-step GBLUP [abstract]. Journal of Dairy Science. 104(Suppl. 1):78(abstr. 200).
Technical Abstract: Official multibreed genomic evaluations for dairy cattle in the US are based on multibreed BLUP evaluation followed by single-breed estimation of SNP effects. Single-step GBLUP (ssGBLUP) allows the straight computation of genomic (G)EBV in a multibreed context. The objective of this study was to develop ssGBLUP multibreed genomic predictions for US dairy cattle. This involved the use of unknown parent groups (UPG) to model the difference in genetic base caused by breed, year of birth, and sex. We used only purebred Ayrshire (AY), Brown Swiss (BS), Guernsey (GU), Holstein (HO), and Jersey (JE). A total of 45M phenotypes for milk (MY), fat (FY), and protein (PY) yields recorded as of January 2020 were available for 19.4M cows. Pedigree information was recorded on 29.5M animals, of which 3.4M were genotyped (Table 1). A 3-trait repeatability model was applied to a complete (reduced) dataset with phenotypes of cows born from 1992 to 2018 (2014). All the effects in the model were breed-specific. Validation for cows was based on correlations between (G)EBV and adjusted phenotypes, whereas for bulls, the latter was replaced by daughter yield deviation. Evaluations were done for each breed separately, AY-BS-GU, and all five breeds together. Reliabilities for bulls and predictability for cows were similar between single-breed and five-breed BLUP. Under ssGBLUP, predictability (reliability) for AY, BS, and GU was on average 21% (9%) lower in the five-breed compared to single-breed model. No changes were observed for HO in the five-breed model because of the greatest number of genotyped animals. Combining AY-BS-GU into one evaluation resulted in predictions similar to the ones from single-breed. Single-step large-scale multibreed evaluations are feasible computationally but fine-tuning is needed to avoid a reduction in reliability when numerically dominant breeds are combined. Possibly, an analysis using an equivalent 60k (or similar) SNP model cannot fully account well for multiple breeds.