Prediction Equations for Gross Energy in Feed Ingredients

The objective of this study was to develop prediction models for gross energy (GE) concentrations in feed ingredients and to evaluate the accuracy of GE prediction equations. A total of 2,279 ingredient samples composed of 58 barley, 319 corn, 13 corn distillers dried grains with solubles, 583 corn gluten feed, 156 copra expellers, 234 copra meal, 12 cottonseed meal, 504 palm kernel expellers, 114 rapeseed meal, 138 soybean meal, 70 wheat, and 78 wheat bran were used. The ingredient samples were analyzed for dry matter (DM), crude protein (CP), ether extract (EE), crude fiber, ash, and GE. The correlation and multiple regression procedures of SAS were used to generate novel prediction equations. Data from 62 ingredients reported in Nutrient Requirements of Swine by the National Research Council (NRC, 2012) were also employed for the validation of the equations. To test the accuracy of GE prediction models, a regression analysis of the measured minus predicted GE concentration against the predicted minus average predicted GE concentration was conducted. The equations developed for estimating GE concentration (kcal/kg on a DM basis) in the feed ingredients are: GE = 4,299 + 7 × CP + 53 × EE (Eq. 1; root mean square error = 206, R 2 = 0.342, and P < 0.001); GE = 4,341 + 11 × CP + 54 × EE – 24 × ash (Eq. 2; root mean square error = 201, R 2 = 0.372, and P < 0.001). All independent variables are in % on a DM basis. A validation test regressing (measured GE – predicted GE) vs. (predicted GE – average of predicted GE) using Eq. 1 and 62 data from NRC (2012) showed that the intercept (−6.9; SE = 46.8 and P = 0.884) and slope (−0.018; SE = 0.089 and P = 0.840) were not different from 0. A validation test of Eq. 2 also showed that the intercept (−56.1; SE = 37.5 and P = 0.140) and slope (0.014; SE = 0.068 and P = 0.834) were not different from 0. The accuracy of an old equation, GE = 4,143 + 56 × EE + 15 × CP – 44 × Ash (Eq. 3), was also tested using 62 data from NRC (2012) and the test resulted in significantly high intercept (101; SE = 35 and P = 0.005) with slope of −0.066 (SE = 0.057 and P = 0.251). The accuracy test of Eq. 3 using 2,279 ingredient data that were used for developing Eq. 1 and 2 also resulted in high intercept (218; SE = 4.3 and P < 0.001) and low slope (−0.121; SE = 0.025 and P < 0.001). Overall, while the old equation appears to underestimate GE concentrations in feed ingredients, the novel equations reported in this work can fairly accurately estimate GE in feed ingredients. Support or Funding Information This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIP) (No. 2016R1A2B2015665).