LAUSR

The objective of this work was to evaluate the precision and accuracy of the milk yield predictions made by the PREP10 model in comparison to those from the National Research Council (NRC) Nutrient Requirements of Dairy Cattle. The PREP10 model is a ration-balancing system that allows protein use efficiency to vary with production level. The model also has advanced AA supply and requirement calculations that enable estimation of AA-allowable milk (MilkAA) based on 10 essential AA. A literature data set of 374 treatment means was collected and used to quantitatively evaluate the estimates of protein-allowable milk (MilkMP) and energy-allowable milk yields from the NRC and PREP10 models. The PREP10 MilkAA prediction was also evaluated, as were both models' estimates of milk based on the most-limiting nutrient or the mean of the estimated milk yields. For most milk estimates compared, the PREP10 model had reduced root mean squared prediction error (RMSPE), improved concordance correlation coefficient, and reduced mean and slope bias in comparison to the NRC model. In particular, utilizing the variable protein use efficiency for milk production notably improved the estimate of MilkMP when compared with NRC. The PREP10 MilkMP estimate had an RMSPE of 18.2% (NRC = 25.7%), concordance correlation coefficient of 0.82% (NRC = 0.64), slope bias of -0.14 kg/kg of predicted milk (NRC = -0.34 kg/kg), and mean bias of -0.63 kg (NRC = -2.85 kg). The PREP10 estimate of MilkAA had slightly elevated RMSPE and mean and slope bias when compared with MilkMP. The PREP10 estimate of MilkAA was not advantageous when compared with MilkMP, likely because AA use efficiency for milk was constant whereas MP use was variable. Future work evaluating variable AA use efficiencies for milk production is likely to improve accuracy and precision of models of allowable milk.