Near-infrared spectroscopy analysis to predict urinary allantoin in dairy cows

Abstract

Accurate quantification of rumen microbial proteins is essential in dairy cow nutrition to estimate the ruminal escape of dietary protein and microbial yield. Current quantification methods rely on indirect measurements using purine derivatives (PD). However, these methods require specialized laboratory equipment and trained personnel, resources which are often not available in farm settings. Near-infrared spectroscopy (NIR) has emerged as a powerful tool for predicting the attributes of biological samples, including meat, corn, soybeans, and liquids. Given that allantoin is the primary component in PD, this study aims to (1) develop a predictive model for allantoin levels in urine using NIR and (2) identify key spectral regions for future applications. A total of 182 urine samples were collected from 182 Holstein cows for colorimetric analysis of allantoin and spectral analysis. The raw spectra were preprocessed using scatter correlation methods and spectral derivatives. The partial least squares regression model achieved an R² of 0.55, a concordance correlation coefficient of 0.73, and a root mean squared error of prediction (RMSEP) of 3.63 mmol/L to predict allantoin concentration from the spectra data set without preprocessing. However, the use of the first derivative (FirstDev) as a preprocessing step reduced the RMSEP from 3.63 mmol/L to 3.25 mmol/L and increased the R² from 0.55 to 0.62. The FirstDev improves spectral resolution by eliminating the constant baseline, potentially explaining the improved model accuracy. Our method has the potential to evaluate the passage rate of microbial protein represented by the changes in urinary allantoin extraction and the potential to be used for AA dietary balance, thereby improving environmental sustainability and profitability in dairy farms.