Prediction of Energy Balance of Friesian x Bunaji dairy Cows using Multivariate Milk Composition Models

The purpose of this study was to investigate whether there is a stable relationship between energy balance (EB) and milk composition measures and if so, how accurate could EB be predicted using group mean data of thirteen (13) primiparous and 47 multiparous (F 1 ) Friesian x Bunaji cows. The energy balance was calculated from changes in body weight (∆BW) and body condition score (∆BCS), and the product of changes in BCS and BW [∆(CS x BW)]. The relationship between EB and milk measures was quantified by Ordinary Least Square (OLS) regression. The eight most informative milk traits that were selected during the exploratory analysis for the development of the models were milk yield (MY), milk protein content (MPC), milk lactose content (MLC), milk protein yield (MPY), protein lactose ratio (PLR), change in milk protein content (dmPc), change in milk lactose content (dmLc) and change in protein lactose ratio (dPLR). These variables were significant (P<0.01) in the exploratory analysis. The models predictive ability was assessed interms of the coefficient of determination adjusted (adj-R 2 ), the root mean square error of prediction (RMSEP) and Akaike’s information criteria (AIC). Using the group mean data, a very high proportion of the variability (R 2 ) in EB (over 90%) was explained by the OLS prediction models in early and mid lactation.  The high percentage of variance explained and the corresponding low prediction error obtained using the OLS regression model clearly indicated that there is a stable biological based relationship between the EB and milk composition measures. Based on the evaluation criteria, it was obvious that the milk composition models worked better in early and mid lactation than late lactation. Also the combinations of the milk composition variables in each of the models varied with stages of lactation. This finding therefore, supports the rationale for developing lactation stages-specific models, for evaluation of energy balance status of dairy cows during lactation.