Protein evaluation in ruminant feeds: metabolisable protein and its limits
Published 22 May 2026 · Updated 10 August 2026
A ruminant is not fed protein; it is fed fermentable substrate and a supply of nitrogen for the microbes that live on it. The protein that reaches the duodenum is a mixture of feed protein that escaped the rumen, microbial protein synthesised in it, and a small endogenous fraction. Crude protein in the feed table says very little about the size or composition of that mixture, which is why crude protein systems were replaced in the 1980s and 1990s by metabolisable protein (MP) systems. The replacements agree on the architecture and disagree, usefully, on the coefficients.
The common structure
Every MP system runs the same four steps. It splits dietary crude protein into a rumen-degradable and an undegradable fraction. It predicts microbial protein synthesis from the energy the rumen ferments. It applies a true-protein share and an intestinal digestibility to both streams. It sums what is absorbed as amino acids. The differences lie in which inputs drive the microbial term and how the undegradable term is corrected for digestibility.
The UK system (AFRC, 1992, and its successors) expresses supply as metabolisable protein, with microbial yield tied to fermentable metabolisable energy and to the effective rumen degradable protein available to match it. The Dutch system (CVB, from the DVE/OEB framework) reports DVE, the truly absorbable protein in the small intestine, together with OEB, the degraded protein balance, a number that says whether the rumen has too little or too much nitrogen relative to fermentable organic matter. The French INRA system uses PDI, with two values per feed, PDIN and PDIE, one defined by nitrogen availability and the other by energy availability; the lower of the two limits microbial output. The US NRC (2001) model carries the same logic with microbial crude protein predicted from total digestible nutrients at about 130 g per kg, and a fixed 80 per cent true protein share.
What the microbes contribute
Microbial protein usually supplies half or more of the amino acids absorbed by a lactating cow, and in low-protein diets the share is higher. Its attraction is the amino acid profile: it is relatively rich in lysine and reasonably supplied with methionine, and it is more consistent between diets than the escape protein of most feeds. That consistency is the reason the amino acid balancing of dairy rations starts from a microbial baseline and adds to it.
The microbial yield terms differ by roughly 10 to 15 per cent among systems. The Dutch and French frameworks use values near 150 g microbial crude protein per kg of fermented organic matter; the US value expressed on a TDN basis lands lower on high-starch diets and higher on forage diets. Within the microbe, 20 to 25 per cent of the nitrogen is nucleic acid, which is absorbed but does not contribute amino acids, and every system has to subtract it somewhere. Some do it in the true protein coefficient, others in the digestibility, and the bookkeeping is not interchangeable.
Where the models disagree
Take a ration of 17 per cent crude protein for a cow eating 24 kg of dry matter. Run it through three systems and the supply of absorbed protein will typically come out within 10 per cent across them, which looks like agreement until the requirements are put alongside. The requirements are also system specific. The efficiency with which absorbed protein is converted to milk protein is set at about 0.67 in the US, UK and Dutch models, but the maintenance requirement and the endogenous loss terms are not the same, so a surplus in one system can be a deficit in another with identical animals and feed.
The sharpest disagreements are in three places. The first is rumen degradability: the rate of passage assumed for the particulate pool alters the calculated escape fraction of a protein meal by a factor that can exceed 0.1 in absolute terms, and passage rate itself changes with intake. The second is intestinal digestibility of escape protein, which is a fixed value in some tables and a feed-specific measurement in others; heat-treated soybean meal, blood meal, and distillers grains do not behave as the average does. The third is the degree to which the system accounts for recycled nitrogen, a pool that matters most when the diet is low in rumen-degradable protein and the animal is supplying urea back to the rumen through saliva and across the wall.
The limits of the whole approach
MP is a protein system, not an amino acid system. The French and US models have gone furthest in adding amino acid profiles for microbial and feed protein, and the INRA 2018 revision extends the PDI framework to individual amino acids, but the coefficients behind those profiles rest on a modest set of duodenal flow studies, with the usual cannulation-related uncertainty in marker recovery. Amino acid absorption and the splanchnic tissues' own use of them is a further correction that few tables carry.
Intake is the other weak point. Microbial protein is predicted from fermented organic matter, and fermented organic matter is a function of intake and passage. Heat load, or any other cause of depressed intake, alters the numbers before the model has been told. Static feed tables cannot follow that; dynamic rumen models can, at the cost of inputs most farms do not have.
For practical formulation, the consistent use of one system matters more than the choice between them. Mixing a French feed value with a Dutch requirement is an error of a larger order than any difference in the underlying biology. Where the milk protein response and the model prediction disagree by more than a few grams per kilogram, the first thing to check is not the model but the feed analysis: the rumen-degradable fraction of the actual silage, not the book one.