The Development and Validation of Prediction Equations for the Apparent Metabolizable Energy of Distillers Dried Grains with Solubles in Broilers
Type of Degreethesis
MetadataShow full item record
The objective of this research was to develop regression equations that accurately predict the AMEn content of distillers dried grains with solubles (DDGS) samples varying in ether extract content and to cross-validate these prediction equations using a set of independent DDGS samples. Experiment 1 determined the nutrient composition and AMEn content of 15 DDGS in order to develop prediction equations for AMEn in broilers. On a DM-basis, AMEn of the DDGS samples ranged from 1,869 to 2,824 kcal/kg. Analyses were conducted to determine gross energy, CP, ether extract, DM, starch, total dietary fiber, neutral detergent fiber, acid detergent fiber, and ash content of the DDGS samples. Stepwise selection resulted in the following best-fit equation for AMEn (DM-basis) based upon the adjusted coefficient of determination (R2adj), root mean square error (RMSE), and prediction error sum of squares (PRESS): AMEn, kcal/kg = -12,282 + (2.60 × gross energy) + (89.75 × CP) + (125.80 × starch) – (40.67 × total dietary fiber) (P ≤ 0.001; R2 = 0.90; R2adj = 0.86; RMSE = 99; PRESS = 199,819). Experiment 2 determined the AMEn content of 15 DDGS in order to validate 4 prediction equations for AMEn of corn DDGS in broilers. On a DM-basis, AMEn of the 15 DDGS samples ranged from 1,975 to 3,634 kcal/kg. Application of the 4 equations to the validation data resulted in RMSE values of 335, 381, 488, and 502 kcal/kg, respectively. The least absolute shrinkage and selection operator technique (LASSO) was applied to proximate analysis data for 30 corn co-products adapted from prior research and resulted in the following best-fit equation: AMEn (kcal/kg) = 3,673 – (121.35 × crude fiber) + (51.29 × ether extract) – (121.08 × ash) (P ≤ 0.001; R2 = 0.70; R2adj = 0.67; RMSE = 457). These results indicated that validation is necessary to accurately determine the risk of error associated with the practical application of prediction equations to external data.