NEAR INFRARED REFLECTANCE SPECTROSCOPY (NIRS) TO ESTIMATE PROTEIN FRACTIONS IN Urochloa GRASS

Abstract

The protein content, after fiber, is a determining factor in the voluntary intake and digestibility of tropical grasses by ruminants. However, its determination in the laboratory is slow and expensive, so it is necessary to develop calibration equations to apply near-infrared reflectance spectroscopy (NIRS) technology in its estimation. The objective was to develop NIRS calibration models to estimate the content of protein fractions in Urochloa sp. grass for the tropical conditions of Mexico. A total of 189 samples of three species of Urochloa grass were obtained, harvested every 35 days for a year and analyzed for the chemical fractions Crude Protein (PC), Non-Protein Nitrogen (NNP), Soluble Protein (PS), Protein insoluble in Neutral Detergent (PIDN), Protein Insoluble in Acid Detergent (PIDA) and the protein nutritional fractions A, B1 + B2, B3 and C were estimated according to the structure of the CNCPS (Cornell Net Carbohydrate and Protein System). The precision of the calibration and validation models were evaluated by the coefficient of determination (R²), the standard error of cross-validation (EEVC), and the residual deviation of the prediction (DRP) obtained by the ratio (DE / EEVC). The validation was carried out by means of a set of 63 external samples. Validation of the equations for PC and PS presented high coefficient of determination (R²) of 0.95 and 0.92, and DRP of 4.0 and 3.4, respectively. The validation for the protein nutritional fractions A and B1 + B2 shows low values of R² (0.69 and
0.86) and DRP (1.6 and 1.9), respectively. The validations of the equations for protein fractions bound to the cell wall [PIDN, PIDA (C) and B3], with R² ≤ 0.39 and DRP ≤ 0.9 were not sufficiently explained. It is concluded that the NIRS calibration models developed to estimate PC and PS are accurate and reliable.