Rapid and nondestructive prediction of total starch and amylose contents in single sorghum kernel (SSK) based on near infrared (NIR)spectroscopy

Authors: ZHOU, JIANJUN - Kansas State University; LI, YONGHUI - Kansas State University; Bean, Scott; ARMSTRONG, PAUL - US Department Of Agriculture (USDA); Wu, Xiaorong

Submitted to: Carbohydrate Polymers
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/10/2025
Publication Date: 8/10/2025
Citation: Zhou, J., Li, Y., Bean, S.R., Armstrong, P., Wu, X. 2025. Rapid and nondestructive prediction of total starch and amylose contents in single sorghum kernel (SSK) based on near infrared (NIR)spectroscopy. Carbohydrate Polymers. 360. Article 124257. https://doi.org/10.1016/j.carbpol.2025.124257.
DOI: https://doi.org/10.1016/j.carbpol.2025.124257

Interpretive Summary: Starch and amylose contents significantly influence sorghum value as a trading commodity and its quality in industrial applications, making their measurement crucial throughout the supply chain, from breeding to end-use applications. This study aimed to develop non-destructive, near-infrared (NIR) spectral methods for quickly estimating amylose and starch contents in individual sorghum kernels. An instrument, designed and built in-house, was used to develop the measurement methods. The instrument is also capable of sorting kernels. Reliable procedures for measuring starch and amylose in single kernels were established, with measurement errors below 1%. Overall, this study successfully developed methods to accurately measure amylose and starch contents in single sorghum kernels and established NIR models to predict these contents, offering a valuable tool for evaluating starch/amylose profiles in sorghum for breeding programs and the selection of seed lines targeting specific agronomic objectives.

Technical Abstract: Starch content and amylose level significantly affect sorghum properties and applications, making their measurement crucial from early breeding stages to commodity trading as well as final industrial applications. This study aimed to establish near-infrared (NIR) spectroscopy models for fast prediction of apparent amylose (AA) and total starch (TS) content in single sorghum kernel (SSK). Reliable procedures for quantifying TS and AA in SSK were established, which achieved high accuracy with test errors below 1.0%. The PLS model with 2 latent variables (LVs) for AA prediction performed well when spectra of the calibration set (251 samples) were preprocessed by normalization + generalized least squares weighting (glsw, a = 0.02), which had a R^2cal of 0.91, R^2cv of 0.85, RMSEC of 1.90% and RMSECV of 2.47%. It showed an R^2pred of 0.83 and RMSEP of 2.58% when validated with the 125 kernel independent sample set. The optimal SSK-TS NIR PLS calibration model was built from 187 calibration sorghum kernels with 10 LVs after preprocessed with the epo/emm filter + mean centering, which had a R2cal of 0.79, R2cv of 0.40, RMSEC of 2.76% and RMSECV of 4.93% and showed a R2pred of 0.72 and RMSEP of 3.19% when applied to an independent validation set of 93 samples. Overall, this study successfully developed wet chemistry methods for measuring AA and TS contents in SSK and established NIR models for nondestructive prediction and sorting of sorghum kernels by their TS or AA content, which will be a useful tool for sorghum breeding as well as sorghum application research.