Extended multiplicative signal correction to improve prediction accuracy of protein content in weathered sorghum grain samples

Abstract

BACKGROUND AND OBJECTIVES: Sorghum grains may be subjected to weathering damage in the field as naked grains on the panicle are exposed to adverse environmental conditions during grain development or dry down before harvest. A near infrared (NIR) protein calibration model did not work as expected for weathered grains. Extended multiplicative signal correction was tested as a preprocessing method for spectra to improve the prediction accuracy of the model when weathered grain samples are evaluated. FINDINGS: NIR spectra of weathered sorghum grain samples showed marked distortions with a higher apparent absorbance at lower wavelengths in the 950–1650 nm spectral range. A partial least squares calibration developed with sound grain samples using spectra preprocessed with multiplicative signal correction (MSC) predicted protein content of validation sample set of sound grains with a root mean square error of prediction (RMSEP) = 0.67% and coefficient of determination (r²) = 0.83. However, performance of this calibration model dropped with RMSEP = 13.14% and r² = 0.04 when used to predict proteins in weathered grain samples. Introduction of weathered grain spectra to the calibration improved predictive performance with RMSEP = 0.92% and r² = 0.69. Application of extended multiplicative scatter correction (EMSC) to preprocess the spectra of the calibration with weathered grain samples enhanced the prediction performance with RMSEP = 0.61% and r² = 0.85. CONCLUSIONS: Inclusion of weathered grain spectra in the calibration and preprocessing spectra with EMSC markedly improved the robustness of the sorghum protein calibration. SIGNIFICANCE AND NOVELTY: Sorghum breeders using NIR spectroscopy for evaluation of protein levels in sorghum breeding lines can get the weathered grain samples also analyzed for protein levels by using this method.

Plant phenotyping relevance

ソルガム穀粒のタンパク質含量をNIRで推定する校正法について、weathering対応の前処理（EMSC）と予測性能を検証しており、植物形質取得法が研究の中心である。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.