Prediction of the chemical composition of pig feces and ileal digesta by near-infrared reflectance spectroscopy (NIRS)

Abstract

Near-infrared reflectance spectroscopy (NIRS) allows the estimation of the chemical composition of various samples (ingredients, animal products, etc.); however, there is little information on its use with samples of swine ileal digesta (ID) or feces (F). Therefore, the objective was to develop prediction equations based on the partial least squares method to predict the chemical composition of the ID and F through NIRS. 110 ID and 202 F samples from digestibility experiments were used, their spectra were obtained and a multivariate model was used to develop the prediction method. The variables analyzed in ID were: crude protein (CP), leucine (Leu), lysine (Lys) and threonine (Thr) and in F: dry matter (DM), CP and energy (E). The DI values were: PC: R2 0.98, standard error of calibration (SEC) 0.330, standard error of prediction (SEP) 0.640; Leu: R2 0.95, SEC 0.040, SEP 0.102; Lys: R2 0.93, SEC 0.077, SEP 0.143; Thr: R2 0.67, SEC 0.209, SEP 0.187. F values were: PC: R2 0.98, SEC 0.95, SEP 1.19; E (kcal/kg): R2 0.94, SEC 60.8, SEP 95.3; MS: R2 0.87, SEC 0.83, SEP 1.15. The results show that the calibration robustness (SD/SEP) was good for PC, 3.34, Leu 2.07 and Lys 2.48 and fair for Thr 1.94, the prediction (RPD) was good for PC 2.11 in ID. In F the R2 were high for PC 0.98 and E 0.94. The highest robustness was for PC 5.59 and its prediction was excellent 4.16 and good for E 2.53. It is concluded that NIRS can predict PC in ID and PC and E in F. In order to improve the estimation of amino acids in ID, the causes affecting the robustness of the calibrations should be explored.