Nutritive evaluation and ingredient prediction of compound feeds for rabbits by near-infrared reflectance spectroscopy (NIRS)

Near-infrared reflectance spectroscopy (MRS) was used to predict the nutritive characteristics of 66 compound rabbit feeds from three countries (Belgium, Spain and Italy) and the main ingredient inclusion rate in 59 of these feeds of known ingredient composition. Principal component analysis (PCA) was performed to classify the compound feeds according to their origin. The coefficient of multiple determination (R-2) for crude protein concentration (CP) was ca. 0.88 in both, calibration and validation with standard errors of calibration (SEC) and prediction (SEP) equal to 7.5 and 7.7 g (kg DM)(-1), respectively. NIRS prediction of gross energy (GE) and digestible energy (DE) concentrations was more precise, with high R-2 (0.90) and low SEP (0.26 and 0.37 MJ (kg DM)(-1), respectively). Satisfactory results were also obtained for both, the dry matter digestibility (DMd) and gross energy digestibility (GEd) prediction. The CP-correlated wavelengths were observed to be associated with the bond vibrations of the protein functional groups, while the wavelengths correlated with GE, DE, DMd and GEd were linked with starch, protein and crude fiber structure. The calibration on absorbance data to estimate the inclusion rate of the main ingredients demonstrated a fair correlation for alfalfa meal, barley and wheat bran, intermediate for sunflower meal and weak for soybean meal. In validation, the precision of the NIRS estimate remained satisfactory for alfalfa and sunflower meal but decreased for barley and wheat bran. The calibration of the spectra transformed in second derivative appeared to improve the quality of estimation by reducing the number of optimal factors from 9-15 to 2-4; moreover, the estimate precision of soybean and sunflower meal inclusions improved (R-2: 0.90 and 0.86, respectively) with the reduction of SEC (13.0 and 12.9 g kg(-1), respectively). In validation, however, the estimate precision for all raw materials became weaker than the degree achieved using absorbance data. PCA on the transformed spectra grouped the compound rabbit feeds according to their country of origin and indicated the possibility of identifying the presence of specific ingredients (i.e. full-fat rapeseed). (C) 1999 Elsevier Science B.V. All rights reserved