| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LETTER TO THE EDITOR |
Objective of the Model.
The objective of the work was to develop a model that would predict N excretion and partition it into fecal, urinary, and milk N, which has huge implications in ever-increasing concerns of nitrate pollution in Europe and North America. Nitrogen excreted from dairy cows contributes to environment pollution through its emissions as nitrogen oxide, nitrous oxide, and ammonia in air and nitrates on soil and water. Dietary manipulation is one of the ways of reducing N excretion from dairy cows and the development of a dynamic model is an important tool in understanding N metabolism in the animal. Therefore, we do not think such a model to have a "restricted objective" as Miller and Baig claimed.
Mechanistic vs Empirical Modeling and Optimization.
Miller and Baig seem to have a different view on what makes a model "mechanistic." In our view, a mechanistic model is based on analysis of underlying components. In this model, it is not at a detailed level, but it is still mechanistic. Mechanistic models seek to understand causation. They are constructed by dividing a system into components and explaining whole-system behavior in terms of those components and their interactions. Miller and Baig’s interpretation seems to be that all known biology must be represented. This model aims to understand how N intake and its subsequent digestion and partition affect emissions to the environment. Therefore, mechanism is at the heart of this model.
Any mechanistic model uses parameter values. Some parameter values are obvious (e.g., 1 mol of glucose cannot provide more than 2 mol of acetic acid). Others have to be estimated based on literature data or simply guessed (e.g., ammonia affinity constants can be estimated from protein-free diets in which the urea level is varied). Therefore, the supposed flaw of mechanistic modeling linked to optimization is incorrect. What Miller and Baig perhaps aim at is that data used for model derivation cannot be used for independent evaluation. However, we used one data set for optimization and another for evaluation. When a parameter is "fitted," it does not indicate the biological relevance of this parameter or otherwise. The model description determines such meaning. Statistical estimates are used widely to estimate biological parameters that have not or could not be determined directly. It follows then that the parameters do have biological meaning. Estimating in whatever way does not change the concept of biology.
Miller and Baig gave an example to make a general point on optimization. On the specific point of varying protein degradability values in model construction, what they failed to understand is that we used the results of five independent experiments; therefore, the initial estimate was only an average value as there were 30 different types of diets considered for model construction. It is elementary logic that for better accuracy of prediction, rumen-undegradable dietary protein values must be given as a fixed value. The model is flexible enough to allow the user to fix the rumen-undegradable dietary protein value if measured or to let the model optimize if it is not measured.
Validity of Data Set.
Detailed information is given in two Journal of Animal Science publications (Castillo et al., 2001a,b); therefore, there was no need to get into much detail. The paper clearly states that the model was developed and evaluated on similar types of animals (Holstein-Friesian) that were in early to mid-lactation with a narrow range of DMI. However, as the main objective of the model was to investigate N metabolism, it was essential to have a wide range of N intakes, which is reflected in the data set. Energy is extremely important in N metabolism (e.g., Castillo et al., 2001a), as shown in the sensitivity analysis of the model, and there are whole sections devoted to this subject in the article. Miller and Baig forget that it has already been stated that the diets were isoenergetic, so there is no need to have a variable parameter. However, in cases where the energy value of the feed is different from the diets used in the model, one can simply change the FME value of the model (E), which will take into account the differences in energy availability. Miller and Baig criticize the model for failing to adequately represent milk production. They again fail to understand the level of complexity of the model. Detailed models have been published that only deal with the mammary gland and AA transport and incorporation to milk (e.g., Hanigan et al., 2001a,b).
"Empiricism" of Estimating Peptides and Endogenous N.
The critics gave examples of "empiricism" of the model by questioning why we let the model estimate the amount of peptides incorporated by microbes in the rumen (YMiDi). We would like to point out again that the level of aggregation for this model is at the whole-animal level, and therefore detailed rumen functions were not included. Such models have been the subject of the co-authors of this article and have been extensively published previously (e.g., Dijkstra et al., 1992). To answer the specific points brought up, the amount of peptides and amino acids incorporated in the rumen by microbes is highly dependent on the diet. Nolan (1993), in his review of N kinetics, states that the rate and extent of dietary protein degradation depends, among other things, on the level of proteolytic activity, which is highly variable and influenced by the diet. The fact that we had different types of diets is reflected in the high standard deviation of the estimate and indicates that the diet had significant effect. There is a wide range of values given in the literature for peptide assimilation, mainly because of diet differences. Several studies have come up with widely varying values for the amount of peptides incorporated by microbes. For example, Miller and Baig quoted Nolan (1975) as reporting 43% in lucerne chaff (in sheep), Beever (1993) reported 35% for grass silage or dried grass, and Chen et al. (1987) puts a maximal limit of 30%. The rumen model of Dijkstra et al. (1992) and its subsequent modifications lets the bacteria vary based on type and amount of soluble protein present by assigning different affinity constants to different groups of bacteria present in the rumen.
The units for endogenous fecal N excretion (NEn) should have been specified as g/g N because they are estimated as the proportion of the metabolic fraction. Once again, the model is criticized for not meeting requirements that are outside its objectives. As with the example of peptides, reports in the literature regarding estimates of endogenous N in ruminants are widely variable and depend on the level of N intake and available energy in the rumen and postrumen. The model had a large standard error for this parameter indicating that more detail is required to improve the parameter estimate. This is an area where more work is required, and an attempt to model endogenous N flow in the gastrointestinal tract of ruminants was made previously (Assis et al., 1997).
It is a gross misunderstanding of the model and a complete disregard of the analysis summarized in Figure 8 for Miller and Baig to state that the model assumes all dietary proteins of whatever source are 65% degraded in the rumen. In fact, one of the major objectives of the model was to assess how protein degradability affects N excretion, and therefore rumen protein availabilities of 50, 60, and 70% were evaluated.
Fecal Microbial N.
The issue of estimating fecal microbial N, and especially fermentation of carbohydrates, in the hindgut is a trade-off between level of system complexity and acceptable level of accuracy in predicting N excretion. In our previous paper (Castillo et al., 2001a), we confirmed the suggestion of Reynolds et al. (1997) that when slow degradable starch sources are used, there is a possibility of increased starch digestion in the postruminal tract, which could stimulate microbial protein synthesis in the cecum and therefore increase fecal output of N at the expense of urinary N (and not milk N).
Energy Coefficients.
Perhaps the energy coefficients were not clearly described in the paper. The principle is that there is a maximal amount of N incorporated by the microbes, which is related to available energy and this is estimated to be 1.76 g N/MJ FME (AFRC, 1993) or 1.6 to 2 g N/MJ (Russell et al., 1992). Assuming an average of 1.8 g N/MJ FME incorporation and a 17 kg DMI/d, in a diet containing 9 MJ FME/kg DM the maximal microbial N incorporation would be 275.4 g/d. In order to keep the ability to change the FME value relatively easy, [E] was defined as the FME value of diet and the maximal incorporation then would be [E] multiplied by DMI (1.8 x 17 = 30.6). The units of kUrMi and KUrMi should be in g N • kg DM/MJ • d and g N, respectively. Perhaps DMI could be explicitly shown in equation [4.3] to make it more clear. In contrast to comments by Miller and Baig, by adjusting the DMI and FME values of the diet, one can predict N incorporation by microbes from the model. However, as N incorporation of 1.8 g/kg FME was derived for lactating cows at 3x maintenance, the basic amount of N incorporation would have to change for cows fed at a higher level of feeding. As shown earlier, if the model can be run for different degradability coefficients and with higher degradable dietary AA input, we would get more microbial protein when it is in short supply. If [E] were lacking, then more AA will not give more microbial AA in the model. This is biologically pretty sound.
Conclusion.
There is a fundamental difference of opinion on model expectations. One cannot criticize a model for not achieving the objectives it did not set out to meet. As Baldwin and Koong (1980) stated, "the objectives of the model should identify the area to be studied, goals of the intended activity, and the level of system to be studied. A model is judged valid when the objectives are attained." One cannot, therefore, judge a model built at the whole-animal level for failing to model the rumen system or postruminal fermentation adequately. As described in the introduction, it is a first step toward describing biological processes in the animal and to evaluating nutritional changes for their potential contribution to environmental pollution. Work is underway to integrate parts of the N model with the detailed mechanistic rumen model of Dijkstra et al. (1992) and the mechanistic methanogenesis model of Mills et al. (2001) with the final aim of being able to predict N and methane losses from the animal.
Literature Cited
AFRC. 1993. Energy and Protein Requirements of Ruminants. Agricultural and Food Research Council. CAB International, Wallingford, UK.
Assis, A. G., J. France, J. Dijkstra, and D. M. Veira. 1997. A model for estimating endogenous protein flows in the gastrointestinal tract of ruminants. J. Anim. Feed Sci. 6:289–301.
Baldwin R. L., and L. J. Koong. 1980. Mathematical modelling in analyses of ruminant digestive function: philosophy, methodology and application. Pages 251–268 in Digestive Physiology and Metabolism in Ruminants: Proceedings of the 5th Int. Symp. Rumin. Physiol. Y. Ruckebusch and P. Thivend, ed. M.T.P. Press, Lancaster, UK.
Beever, D. E. 1993. Rumen function. Pages 187–215 in Quantitative Aspects of Ruminant Digestion and Metabolism. J. M. Forbes and J. France, ed. CAB International, Wallingford, UK.
Castillo, A. R., E. Kebreab, D. E. Beever, J. H. Barbi, J. D. Sutton, H. C. Kirby, and J. France. 2001a. The effect of energy supplementation on nitrogen utilization in grass silage diets by lactating dairy cows. J. Anim. Sci. 79:240–246.
Castillo, A. R., E. Kebreab, D. E. Beever, J. H. Barbi, J. D. Sutton, H. C. Kirby, and J. France. 2001b. The effect of protein supplementation on nitrogen utilization in grass silage diets by lactating dairy cows. J. Anim. Sci. 79:247–253.
Chen, G., J. B. Russell, and C. J. Sniffen. 1987. A procedure for measuring peptides in rumen fluid and evidence that peptide uptake can be a rate-limiting step in ruminal protein degradation. J. Dairy Sci. 70:1211–1219.
Dijkstra J., H. D. Neal, D. E. Beever, and J. France. 1992. Simulation of nutrient digestion, absorption and outflow in the rumen: Model description. J. Nutr. 122:2239–2256.
Hanigan M. D., B. J. Bequette, L. A. Crompton, and J. France. 2001. Modeling mammary amino acid metabolism. Livest. Prod. Sci. 70:63–78.
Hanigan M. D., L. A. Crompton, J. A. Metcalf, and J. France. 2001. Modeling mammary metabolism in the dairy cow to predict milk constituent yield, with emphasis on amino acid metabolism and milk protein production: Model construction. J. Theor. Biol. 213:223–239.[Medline]
Kebreab, E., J. France, J. A. N. Mills, R. Allison, and J. Dijkstra. 2002. A dynamic model of N metabolism in the lactating dairy cow and an assessment of impact of N excretion on the environment. J. Anim. Sci. 80:248–259.
Miller, E. L., and M. Y. Baig. 2002. Critique of a dynamic model of N metabolism in the lactating dairy cow.
Mills, J. A. N., J. Dijkstra, A. Bannink, S. B. Cammell, E. Kebreab, and J. France. 2001. A mechanistic model of whole-tract digestion and methanogenesis in the lactating dairy cow: Model development, evaluation, and application. J. Anim. Sci. 79:1584–1597.
Nolan, J. V. 1975. Quantitative models of nitrogen metabolism in sheep. Pages 416–431 in Digestion and Metabolism in the Ruminant. University of New England, Armidale, NSW, Australia.
Nolan, J. V. 1993. Nitrogen Kinetics. Pages 123–143 in Quantitative Aspects of Ruminant Digestion and Metabolism. J. M. Forbes and J. France, ed. CAB International, Wallingford, UK.
Reynolds, C. K., J. D. Sutton, and D. E. Beever. 1997. Effects of feeding starch to dairy cattle on nutrient availability and production. Pages 105–134 in Recent advances in animal nutrition. P. C. Garnsworthy and J. Wiseman, ed. Nottingham Univ. Press, UK.
Russell, J. B., J. D. O’Connor, D. G. Fox, P. J. Van Soest, and C. J. Sniffen. 1992. A net carbohydrate and protein system for evaluating cattle diets: I. Ruminal fermentation. J. Anim. Sci. 70:3551–3561.[Abstract]
| ||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
