Genomic Prediction of Genetic Residual Feed Intake Integrating a Novel Energy Sink for Change in Body Reserves.

Abstract

Traditionally, a two-step modeling approach of residual feed intake (RFI) is incorporated into the Feed Saved index at dairy cattle genetic evaluation centers. Challenges have been identified in the 1^st step on handling fixed effects in the statistical model and dealing with missing phenotypes. This could be solved using a multi-variate modelling approach for genetic RFI (gRFI). Most existing RFI models use changes in body weight, and therefore, likely inadequately account for changes in body reserves because energy density differs between mobilization and deposition, and between adipose and muscle tissue. Alternatively, energy balance can be estimated from body reserve changes (EB_body). Therefore, this study aimed to explore a genomic evaluation of gRFI in Nordic primiparous cows using EB_body as energy sink for changes in body reserves.

                      Weekly records were collected from 2,029 Jersey (JER) cows, 3,178 Red Dairy Cattle (RDC) cows, and 4,661 Holstein (HOL) cows. For JER and RDC, the feed intake data was obtained with the Cattle Feed InTake system (CFIT, VikingGenetics, Denmark). For HOL, feed intake data was collected from CFIT farms and a research farm (857 cows and 25,547 weekly records). The genotyping rate for cows with data were 92% for JER and RDC, and 81% for HOL.

                      The gRFI model was a random regression multi-variate model with 2^nd order Legendre polynomials for additive genetic and permanent environmental effects. The gRFI model was validated with an across-herd cross-validation scheme using the Legarra Reverter method and reporting bias, dispersion and correlation terms. Breeding values were predicted using the single-step approach for both genotyped and non-genotyped animals. The bias was close to 0 for all breeds. The dispersion coefficients were found in an acceptable range at 0.92 (DMI) and 0.87 (gRFI) for HOL and 0.96 (DMI) and 0.85(gRFI) for RDC, while overdispersion was observed for JER (DMI:0.75, gRFI:0.69). Correlations between genomic breeding values, estimated with whole and partial phenotypic information, were moderately high for all breeds (DMI: 0.51-0.68, gRFI: 0.46-0.59). In conclusion, it was possible to construct a genomic gRFI model for all three Nordic dairy cattle breeds and integrate EB_body as an energy sink indicator. We observed promising validation metrics for HOL and RDC, but JER models need further refinement. The results demonstrate selection for gRFI is expected to provide genetic gain of feed efficiency in dairy cattle.