Modelling equine pre-cecal digestible crude protein using feedstuff crude protein and fiber composition

Protein digestion and the absorption of amino acids and small peptides occurs in the small intestine of the horse. Current NRC protein requirements are based on crude protein (CP); however, distribution of protein in feedstuffs between cell contents and cell wall affects protein digestion in the foregut. Equine nutrient requirement models use a widely available laboratory assay, neutral detergent soluble crude protein (NDSCP) to quantify pre-cecal digestible crude protein (pcdCP), a predictor to determine if protein requirements are met. Although pcdCP can be determined in the laboratory, development of linear regression models based on more common assays are potentially useful for predicting protein availability in the foregut at a lower cost. Therefore, our study objective was to develop linear regression models to predict pcdCP using CP and fiber composition in feedstuffs common to equine diets. Data were collected from the 2023 cumulative Dairy One© Feedstuff Composition Library. Feedstuffs were selected based on having an n ≥50 for each individual nutrient analysis used for model development (CP, neutral detergent insoluble crude protein [NDICP], neutral detergent fiber [NDF], acid detergent fiber, acid detergent lignin). Feedstuff NDSCP was calculated as NDSCP = CP − NDICP, and pcdCP was calculated as pcdCP = 0.9 × NDSCP. Feedstuffs were categorized into 4 groups (CAT): forages; whole and processed grains; grain byproducts; oilseeds and oilseed meals. Data were analyzed using R (v4.4.0). Model variable selection of feedstuff nutrients was performed using a Pearson correlation. Models were ranked using the Akaike information criterion (AICc) and AICc weights (AICcWt). Model adjusted (adj.) R² was used to determine explanatory power of each model. Significance for a variable slope was set at P ≤ 0.05. The model using CP+NDF (pcdCP [g/kg DM] = 5.654 + 0.7615 × CP − 0.0313 × NDF) was the highest-ranked model (AICcWt = 0.95; adj. R² = 0.987). Accounting for categorical differences (CP+NDF+CAT) decreased explanatory power (adj. R² = 0.986) and predictor power (AICcWt = 0.05). Furthermore, categorical adjustments were not significant (P > 0.15) and likely accounted for in the base model using NDF. No other models were within the threshold of ΔAICc ≤10.0. Overall, pcdCP could be predicted with high precision using only CP and NDF as predictor variables, both of which have been used to predict CP digestibility. These results indicate that more precise estimates of small intestine absorbable protein supply may be attained through further model development and evaluation of protein fractionation in feedstuffs.