What question did this study set out to answer?

The research aims to develop a NIR spectroscopy model for accurate prediction of β-glucan content in naked barley grains.

March 29, 2026Open Access

Establishment of a near-infrared reflectance spectroscopy model for predicting β-glucan content in naked barley grain

Key Points

The research aims to develop a NIR spectroscopy model for accurate prediction of β-glucan content in naked barley grains.
Collected NIR spectra from 215 naked barley grain samples
Used enzymatic methods to determine β-glucan content
Applied 14 preprocessing techniques for spectral data
Employed PLSR, SVM, and PCR for model construction
Evaluated six feature variable selection methods for model optimization
The SVM model with second derivative preprocessing showed the best predictive performance
CARS-SD-SVM model achieved highest accuracy with R² values of 0.859 and 0.877
RMSEC of 0.272 and RMSEP of 0.237 indicate low prediction errors
RPD value of 2.790 demonstrates excellent prediction capability

Abstract

β-glucan, known for its roles in regulating blood sugar, lowering blood lipids, and exhibiting anti-tumor and antioxidant properties, is one of the most important quality indicators of naked barley. Establishing a near-infrared (NIR) prediction model for β-glucan content enables rapid and non-destructive evaluation of this trait in early-generation naked barley materials, which can significantly enhance breeding efficiency. In this study, NIR spectra were collected from 215 naked barley grain samples, and β-glucan content was determined using enzymatic methods. Based on spectral scanning and Monte Carlo cross-validation, 10 outlier samples were excluded. The remaining samples were divided into calibration and validation sets using the SPXY method. Fourteen preprocessing techniques, including normalization, first derivative, and second derivative, were applied to the spectral data. Partial least squares regression (PLSR), support vector machine (SVM), and principal component regression (PCR) were employed to construct prediction models for β-glucan content. Among these, the SVM model combined with the second derivative preprocessing showed the best predictive performance. To further optimize this model, six feature variable selection methods—successive projections algorithm (SPA), uninformative variable elimination (UVE), competitive adaptive reweighted sampling (CARS), variable iterative space shrinkage algorithm (VISSA), bootstrapping soft shrinkage (BOSS), and iteratively retained informative variables (IRIV) were evaluated. The CARS-SD-SVM model achieved the highest accuracy, with R2C, RMSEC, R2P, RMSEP, and RPD values of 0.859, 0.272, 0.877, 0.237, and 2.790, respectively. This model not only demonstrated excellent predictive performance but also high reliability, making it suitable for the rapid analysis and prediction of β-glucan content in naked barley grains. It holds great potential for application in naked barley breeding and quality assessment.

AI से पूछें

Bookmark

View Full Paper