Abstract #T48
Section: Breeding and Genetics
Session: Breeding and Genetics II
Format: Poster
Day/Time: Tuesday 7:30 AM–9:30 AM
Location: Exhibit Hall B
Session: Breeding and Genetics II
Format: Poster
Day/Time: Tuesday 7:30 AM–9:30 AM
Location: Exhibit Hall B
# T48
Genomic prediction of lactation curves for milk, fat, protein, and somatic cell score in Canadian Jersey cattle.
H. R. Oliveira*1,2, L. F. Brito1, J. Jamrozik3,1, F. F. Silva2, F. S. Schenkel1, 1University of Guelph, Guelph, ON, Canada, 2Universidade Federal de Viçosa, Viçosa, Minas Gerais, Brazil, 3Canadian Dairy Network, Guelph, ON, Canada.
Key Words: GBLUP, genome-wide selection, random regression
Genomic prediction of lactation curves for milk, fat, protein, and somatic cell score in Canadian Jersey cattle.
H. R. Oliveira*1,2, L. F. Brito1, J. Jamrozik3,1, F. F. Silva2, F. S. Schenkel1, 1University of Guelph, Guelph, ON, Canada, 2Universidade Federal de Viçosa, Viçosa, Minas Gerais, Brazil, 3Canadian Dairy Network, Guelph, ON, Canada.
Application of random regression models (RRM) in 2-step genome-wide selection (GWS) may provide opportunities for selecting young animals based on the pattern of the lactation curve, without changing the traditional genetic evaluation system used in several countries. In this context, the prediction accuracy of direct genomic values (DGVs) for milk (MY), fat (FY) and protein (PY) yields, and somatic cell score (SCS) over days-in-milk in a 2-step genomic evaluation approach was investigated. Estimated breeding values for each test-day (from 5 to 305 d) from the first 3 lactations of Jersey cows (referred as 1, 2 or 3 beside trait acronyms), derived from estimates of the lactation curve coefficients (Legendre polynomials of order 4), were de-regressed (dEBVs) and used as pseudo-phenotypes in the second step of GWS. Genotyped individuals included in the official Canadian Jersey genetic evaluation in December, 2012, by the Canadian Dairy Network (CDN; Guelph, ON, Canada) were used as training population (n = 1,463 animals). The validation population included 315 individuals born after 2012, which had an official genetic evaluation in December, 2016. Individual additive genetic random regression coefficients for each trait were predicted using Genomic Best Linear Unbiased Prediction (GBLUP) and further used to derive DGV for each day of the 305d lactation. Prediction accuracy for each trait was evaluated based on Pearson correlation between DGVs and dEBVs (rDGV,dEBV) adjusted for the average reliability of dEBVs in the validation population. The average estimated rDGV,dEBV over the lactation curve was 0.64, 0.73, and 0.75 for MY1, MY2, and MY3; 0.53, 0.50, and 0.54 for FY1, FY2, and FY3; 0.85, 0.74 and 0.56 for PY1, PY2, and PY3; and 0.26, 0.54, and 0.37 for SCS1, SCS2, and SCS3, respectively. Therefore, the use of RRM in 2-step GWS produced moderately accurate DGVs for milk production traits and SCS over the lactation in Canadian Jersey cattle. Strategies to optimally blend DGVs and traditional RRM EBVs will be investigated next.
Key Words: GBLUP, genome-wide selection, random regression
