Describing aNDFom in vitro digestion with a multi-compartment model and evaluation of predictions in the CNCPS v7.0 model

• Published in: Journal of animal science Vol. 94; p. 802
• Main Authors: Zontini, A M, Van Amburgh, M E
• Format: Journal Article
• Language: English
• Published: Champaign Oxford University Press 01-10-2016
• Subjects: Byproducts; Cattle; Diet; Differential equations; Digestibility; Digestion; Digestive system; Experimental design; Lag phase; Mathematical models; Milk; Predictions; Proteins; Regression analysis; Statistical analysis; Statistical models; Studies; Substrates
• ISSN: 0021-8812; 1525-3163
• DOI: 10.2527/jam2016-1646

Models such as the CNCPS rely on having accurate values for digestion rate to aid in predicting metabolizable energy and protein yield from the rumen and subsequent milk yield and total tract digestibility. In vitro aNDFom digestion displays a sigmoidal shape, which is the result of a two-step process: the lag phase, where bacteria attach to substrate and establish digestion and the digestion phase where substrate is being degraded. In previous mechanistic models, the lag phase was discrete, meaning that aNDFom digestion could occur only after the lag phase was over. In other statistical models aNDFom digestion has been described with a multi-compartment system using γ functions. Such models can describe the sigmoidal shape of aNDFom digestion but the estimated rate of digestion were probabilistic and not deterministic. The objective of this study was to develop a mechanistic model that describes the sigmoidal behavior of in vitro aNDFom digestion, using differential equations to provide deterministic values of digestion rates. The proposed are multi-compartments models (Fig. 1), one describing aNDFom in plant by-products, where the digestible aNDFom fraction (dNDF) is unique; and the other describing aNDFom digestion in forages, where dNDF is further fractionated into fast and slow digesting pools. In both feed types a fraction remains un-digested over time (uNDF). In these models, the lag phase is a rate, the life-time of the dNDF fractions is consistent with a γ distribution, and the behavior of the system is sigmoidal. The model was fitted to 36 conserved forages, 32 fresh forages; and 15 plant byproducts analyzed in-vitro. The quality of fit was evaluated with overall slope (1.03), intercept (0.01), R2 (0.98), and an RMSE (0.02) of the regressions of observed versus predicted. Further, the relevance of the model predictions were assessed by evaluating the RMSE of CNCPS predictions of ME allowable milk and aNDFom TTD, using information from a lactating cattle study where treatment diets were formulated to quantify the effects of aNDFom source and digestibility. Parameters of aNDFom digestion were calculated for each ingredient of the diets, and used as inputs for CNCPSv7.0. The experimental design was a 3 x 3 Latin square with 21 d adjustment and 5 d sampling periods this was used to develop the RMSE calculations. The RMSE of predictions among the three treatments were 0.8 Kg for ME allowable milk and 4% for aNDFom total tract digestibility.