Select
Application of Near-Infrared Spectroscopy in Amino Acid Analysis of Total Mixed Rations
WANG Yating, LUO Yaping, WANG Chengjie
2025, 46(5):
18-31.
doi:10.12160/j.issn.1672-5190.2025.05.003
Abstract
(
11 )
PDF (5249KB)
(
3
)
Save
References |
Related Articles |
Metrics
[Objective] To investigate the application of near-infrared spectroscopy technology in the analysis of amino acids in total mixed rations (TMR), aiming to meet the market demand for rapid and non-destructive detection of TMR nutritional components. [Methods] Near-infrared spectroscopy technology combined with partial least squares (PLS) regression was utilized to establish near-infrared spectroscopy prediction models for the content of 17 amino acids, crude protein (CP), and dry matter (DM) using 282 dairy cow TMR samples collected from Ningxia, Yunnan, Beijing, Heilongjiang, and Inner Mongolia. [Results] The coefficients of determination for the validation set (Rp 2 ) for arginine (ARG), aspartic acid (ASP), cysteine (CYS), glutamic acid (GLU), glycine (GLY), histidine (HIS), isoleucine (ILE), leucine (LEU), lysine (LYS), phenylalanine (PHE), serine (SER), threonine (THR), tyrosine (TYR), valine (VAL), crude protein (CP), and dry matter (DM) were 0.733, 0.820, 0.852, 0.920, 0.888, 0.726, 0.880, 0.877, 0.900, 0.815, 0.865, 0.824, 0.807, 0.809, 0.906, and 0.905, respectively, all greater than 0.64; the standard error of prediction (SEP ) and systematic bias (Bias ) for the validation set were all less than 0.48. The relative prediction deviation (RPD ) ranged from 2.072 to 3.564. The Rp 2 for methionine (MET), proline (PRO), and alanine (ALA) were 0.584, 0.507, and 0.314, respectively, all less than 0.64; their RPD values were below 1.75. [Conclusion] The established models in this study satisfy the requirements for prediction accuracy for routine analysis of most of the above nutritional components analyzed (ASP, CYS, GLU, GLY, ILE, LEU, LYS, PHE, SER, THR, ARG, HIS, TYR, VAL, CP, DM). The models for certain amino acids (MET, PRO, ALA) require further improvement in sensitivity by optimizing the spectral feature extraction algorithm or expanding the sample size.