Evaluation of the National Academies of Sciences, Engineering, and Medicine (NASEM) milk protein yield prediction model with data from Brazilian commercial farms.

The National Academies of Sciences, Engineering, and Medicine (NASEM, 2021) milk protein yield (MPY) prediction equation includes independent and additive effects of digestible energy intake and absorbed EAA. Our objective was to evaluate the NASEM MPY prediction and EAA use efficiency in Holstein cows in pens from commercial farms. Data collected from 12 Brazilian herds were used. All cows were housed in a freestall or compost barn and fed TMR. For each of the 89 pens (a total of 8,345 cows, 50-325 cows per pen), data on milk production and composition, DMI, DIM, parity, BW, and diet composition were compiled. Data from each pen were entered in NASEM software to predict MPY and efficiency of utilization for each EAA. Pens were divided by observed MPY levels in 3 clusters: low = 970, medium = 1,196, and high = 1,524 g/d MPY, representing the mean values for each cluster. Within each cluster, NASEM MPY prediction was compared with the observed MPY using the coefficient of determination (R), root mean square error (RMSE), and the concordance correlation coefficient (CCC). The MIXED procedure of SAS with the fixed effect of cluster and the random effects of farm and pen nested within farm was used to compare the number of protein sources used in the diets and EAA efficiency by cluster. Overall prediction performance of the NASEM MPY equation was best for the low MPY cluster relative to medium and high ones (CCC = 0.73, 0.37, and 0.35, respectively), with high accuracy (RMSE = 62.9 g/d, 6.5% of the mean) and moderate precision (R = 0.57). On the other hand, despite lower precision (R = 0.39), accuracy was also high for the medium cluster (RMSE = 95.6 g/d, 8% of the mean). Finally, prediction for the high MPY cluster had the highest precision (R = 0.74), but the lowest accuracy (RMSE = 224.7 g/d, 14.7% of the mean). The number of protein sources in the diets was greater in the high and medium productions clusters compared with the low production cluster (4.1, 3.9, and 3.0 sources, respectively; SEM = 0.33). Increasing the production level of the cluster linearly increased the EAA use efficiency of all EAA. The greater pull effect in the higher production groups and the better combination of AA from more protein sources could explain better AA efficiencies.