40
Quantifying variable amino acid efficiencies in lactating sows
This study estimated new and variable individual AA efficiencies that could be incorporated into diet formulation for lactating sows. Literature used to estimate AA efficiencies in the NRC 2012 swine model were mined for feed intake, sow weight loss, dietary N, litter weight gain, litter size, milk yield, and milk AA composition data. Models were derived to estimate variable AA use efficiencies in milk. A mixed-effect regression of AA efficiencies on diet composition, feed intake, and weight loss was conducted accounting a random study effect. Models were evaluated using the root mean squared error (RMSE), concordance correlation coefficient (CCC), and corrected Akaike Information Criteria. Arginine, Lys, Phe, Thr, and Val efficiencies decreased (P < 0.05) with increasing dietary N and all AA efficiencies increased (P < 0.05) with increasing milk yield. Only Leu and Met efficiencies decreased (P < 0.05) with increasing weight loss. Equations predicted measured AA efficiencies with RMSE between 5.3 and 15.4% and CCC between 0.97 and 0.99. The AA efficiencies were also modeled as a function of dietary N, feed intake, weight loss, and efficiencies of Leu, Met, and Cys (these explanatory AA were selected based on their individual fit for direct prediction). Histidine, Ile, and Lys efficiency models improved with Leu and Cys efficiencies included as covariates but models of Arg, Phe, Thr, or Val efficiency did not improve. Energy allowable milk was calculated based on the difference between energy supply and maintenance requirements and the lactation energy use efficiency. Amino acid allowable milk was calculated from modeled AA efficiencies and SID AA supply. The minimum of energy and AA allowable milk estimates and their mean were calculated. A first-limiting nutrient model (minimum of energy and AA allowable milk) had an RMSE of 110%. A co-limiting nutrient model (mean of energy and AA allowable milk) had a lower (32%) RMSE. To more fully investigate the utility of a co-limiting nutrient model, a multi-substrate Michaelis-Menten equation was fit to predict milk yield as a function of AA and ME supply. After a stepwise elimination of nonsignificant parameters, the final model (based on Arg, Leu, Met, Phe, Thr and Val SID intake) returned a RMSE of 10% and very good concordance (0.77). These results suggest that nutrients co-limit milk production and moving toward a more response-driven model may help define more precise diets that account for dynamic mammary uptake of AA.
Keywords: Variable Efficiency; Amino Acid; Lactating Sow