Adsa Logo White Adsa Title White

Prediction of dry matter intake using linear regression of sensor, blood metabolite, and performance variables in mid-lactation cows.

M. J. Martin

Events

06-22-2020

Join M. J. Martin on this page for a live text chat!
6:00 PM - 8:00 PM GMT

Abstract:

M121
Prediction of dry matter intake using linear regression of sensor, blood metabolite, and performance variables in mid-lactation cows.
M. J. Martin*1, R. S. Pralle1, R. L. Wallace2, M. R. Borchers2, S. R. DeNise2, K. A. Weigel1, H. M. White1. 1University of Wisconsin Madison Madison, WI, 2Zoetis Kalamazoo, MI.

Development of useful predictions of dry matter intake (DMI) for individual cows on farms could allow for the assessment of the impact of nutritional interventions on feed efficiency and facilitate the inclusion of thousands of additional cows in genomic reference populations for feed efficiency. The objectives of this study were to develop and evaluate DMI prediction models using cow performance, blood metabolite, and sensor data. Mid-lactation primi- and multiparous Holstein cows (n = 62/replicate) fitted with SMARTBOW eartags were housed in a freestall pen with Insentec feeders (2 replications; 45d). Sensor data collected via SMARTBOW included time spent lying (LT), ruminating (RT), and standing at the feedbunk (FT), as well as activity classified as high-active, active, and inactive. Other data collected included metabolic body weight (MBW), BCS, milk yield (MY), milk energy yield (MilkE), as well as plasma glucose, nonesterified fatty acids (NEFA), β-hydroxybutrate, and triglycerides (TG). All variables were scaled using the full data set and used as predictors of scaled DMI in multiple linear regression models. A model search was performed via the dredge function of MuMIN in R (v. 3.6.2). A 5-fold cross-validation was used to evaluate prospective models based on RMSE, r2, and concordance correlation coefficient (CCC). Variables selected by all top performing models included BCS, MBW, and MilkE. The best performing model included BCS, MBW, MilkE, RT, active time, and NEFA (RMSE = 0.49; r2 = 0.85; CCC = 0.87). Models excluding metabolites included BCS, MBW, MilkE, RT, and active time variables and achieved similar performance (RMSE = 0.49; r2 = 0.84; CCC = 0.87). Performance was slightly poorer for models excluding metabolites and milk composition data; the selected model included BCS, MBW, MY, RT, and active time (RMSE = 0.52; r2 = 0.82; CCC = 0.85). Models using only sensor and MilkE data included MilkE, FT, and LT (RMSE = 0.68; r2 = 0.57; CCC = 0.71). In conclusion, DMI can be reasonably predicted using sensor and performance variables with or without the use of blood metabolites.

Keywords: predictive model, feed efficiency.