Predicting the digestible energy of corn determined with growing swine from nutrient composition and cross-species measurements

The DE values of corn grain for pigs will differ among corn sources. More accurate prediction of DE may improve diet formulation and reduce diet cost. Corn grain sources (n = 83) were assayed with growing swine (20 kg) in DE experiments with total collection of feces, with 3-wk-old broiler chick in nitrogen-corrected apparent ME (AME_N) trials and with cecectomized adult roosters in nitrogen-corrected true ME (TME_N) studies. Additional AME_N data for the corn grain source set was generated based on an existing near-infrared transmittance prediction model (nearinfrared transmittance-predicted AME_N [NIT-AME_N]). Corn source nutrient composition was determined by wet chemistry methods. These data were then used to 1) test the accuracy of predicting swine DE of individual corn sources based on available literature equations and nutrient composition and 2) develop models for predicting DE of sources from nutrient composition and the cross-species information gathered above (AME_N, NIT-AME_N, and TME_N). The overall measured DE, AME_N, NIT-AME_N, and TME_N values were 4,105 ± 11, 4,006 ± 10, 4,004 ± 10, and 4,086 ± 12 kcal/kg DM, respectively. Prediction models were developed using 80% of the corn grain sources; the remaining 20% was reserved for validation of the developed prediction equation. Literature equations based on nutrient composition proved imprecise for predicting corn DE; the root mean square error of prediction ranged from 105 to 331 kcal/kg, an equivalent of 2.6 to 8.8% error. Yet among the corn composition traits, 4-variable models developed in the current study provided adequate prediction of DE (model R² ranging from 0.76 to 0.79 and root mean square error [RMSE] of 50 kcal/kg). When prediction equations were tested using the validation set, these models had a 1 to 1.2% error of prediction. Simple linear equations from AME_N, NIT-AME_N, or TME_N provided an accurate prediction of DE for individual sources (R² ranged from 0.65 to 0.73 and RMSE ranged from 50 to 61 kcal/kg). Percentage error of prediction based on the validation data set was greater (1.4%) for the TME_N model than for the NIT-AME_N or AME_N models (1 and 1.2%, respectively), indicating that swine DE values could be accurately predicted by using AME_N or NIT-AME_N. In conclusion, regression equations developed from broiler measurements or from analyzed nutrient composition proved adequate to reliably predict the DE of commercially available corn hybrids for growing pigs.