Can milk fat changes be predicted from nutrient flows in dairy cows? Design and evaluation of an empirical model - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Livestock Science Année : 2014

Can milk fat changes be predicted from nutrient flows in dairy cows? Design and evaluation of an empirical model

Résumé

Most of the current feed evaluation systems used to estimate dairy cow requirements and formulate diets cannot predict the effects of dietary changes on milk composition, in particular on milk fat content (MFC) and milk fat yield (MFY) responses. Dietary changes alter dairy cow MFC and MFY through modifications in the supply of nutrients, which act as precursors or inhibitors of mammary fat synthesis. Taking into account the supply of nutrients could allow feed evaluation systems to predict milk composition responses to dietary changes. We described an empirical model based on nutrient flows to predict MFC and MFY responses to dietary changes. This model was built by coupling a set of published empirical equations estimating nutrient flows (acetate, propionate, butyrate, glucose and digestible protein) from dietary characteristics and the measured ruminal VFA proportions to another set of equations estimating the response of milk fat to the supply of these nutrients. The model was evaluated by comparing the predicted and observed MFC and MFY responses on three databases derived from published feeding studies in dairy cows. The databases compiled published trials involving a change in the forage-to-concentrate ratio (‘F:C ratio’ sub-database), a change in the carbohydrate source (‘CHO source’ sub-database) or a change in both F:C ratio and CHO source (‘ratio and source’ sub-database). Overall, the current version of the model did not accurately predict milk fat responses: the model did not enable to explain the MFC and MFY responses in the ‘CHO source’ sub-database (P>0.10 and R2=0.02); in the ‘F:C ratio’ sub-database, agreement between observed and predicted responses was better (the slopes were not different from unity and were respectively 0.84 and 0.94 for MFC and MFY), but R2 were low (R2=0.36 for MFC and R2=0.43 for MFY). In the ‘ratio and source’ database, MFY was quite correctly predicted (the slope was 0.93 and R2=0.42), but not the MFC responses (P>0.10 and R2=0.01). The mean errors of prediction for MFC were 2.61, 3.79 and 3.55 g/kg, respectively in the ‘F:C ratio’, ‘CHO source’ and the ‘ratio and source’ sub-databases. The effects of several interfering factors related to cows, diets or experimental conditions were tested to explain differences between observed and predicted responses in each sub-database. They enabled to identify several potential sources of bias in the model: especially, bias in estimation of OM digestibility, bias in estimation of glucose effects, lack of the effects of fatty acids in the model. These constitute interesting ways to improve the model.

Domaines

Biologie animale
Fichier non déposé

Dates et versions

hal-01210581 , version 1 (02-10-2015)

Identifiants

Citer

Gaëlle Maxin, Jean-Louis Peyraud, Pierre Noziere, Henri Rulquin, Frederic Glasser. Can milk fat changes be predicted from nutrient flows in dairy cows? Design and evaluation of an empirical model. Livestock Science, 2014, 164, pp.46-56. ⟨10.1016/j.livsci.2014.03.006⟩. ⟨hal-01210581⟩
91 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More