Mathematical prediction of ileal energy and protein digestibility in broilers using multivariate data analysis

A proportional mixture design was used to systematically create a total of 56 diets using ten feed ingredients. Diets differed widely with regards to chemical characteristics and ingredient inclusion levels. Apparent ileal digestibility of energy and protein of the diets were determined in broiler growers fed ad libitum from 21 to 24 d post-hatch. The chemical composition and the in vivo digestibility values were used to establish prediction equations for energy and protein digestibility, using multivariate data analysis. Root mean square error as percentage of the observed means (RMSE%) and residual error were used to evaluate the strength and accuracy of the predictions and to compare predictions based on chemical characteristics with estimates based on table values. The estimates of ileal digestibility of energy from table values were relatively accurate (RMSE% = 5.15) and was comparable to those predicted based on the chemical composition of diets. Estimates of ileal digestibility of protein based on table values were less accurate (RMSE% = 8.21); however, the prediction was improved by multivariate regression (RMSE% = 5.46) based on chemical composition of diets. The best predictors for ileal energy digestibility were starch, crude fiber and phytate contents (P < 0.01) and the best predictors for crude protein digestibility were starch, CF and fat contents (P < 0.05). In conclusion, the ileal digestibility of energy can be accurately predicted using table values; however, the accuracy of prediction of the ileal digestibility of protein can be improved when chemical characteristics of the diet are considered.