Premium
Moisture effects on robustness of sorghum grain protein near‐infrared spectroscopy calibration
Author(s) -
Peiris Kamaranga H. S.,
Bean Scott R.,
Chiluwal Anuj,
Perumal Ramasamy,
Jagadish S. V. Krishna
Publication year - 2019
Publication title -
cereal chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.558
H-Index - 100
eISSN - 1943-3638
pISSN - 0009-0352
DOI - 10.1002/cche.10164
Subject(s) - moisture , water content , calibration , sorghum , near infrared spectroscopy , robustness (evolution) , chemistry , coefficient of variation , analytical chemistry (journal) , mean squared error , spectroscopy , soil science , biological system , environmental science , agronomy , mathematics , statistics , chromatography , biology , optics , geology , physics , biochemistry , geotechnical engineering , organic chemistry , quantum mechanics , gene
Background and objectives A near‐infrared (NIR) spectroscopy method was developed for rapid and nondestructive evaluation of protein content of intact sorghum grains. Effect of grain sample moisture variation on the robustness of protein calibration was investigated. Findings An initial NIR protein calibration model with sorghum grains in 7.28%–11.57% moisture content range with coefficient of determination ( R 2 ) of 0.94 and standard error of cross‐validation (SECV) of 0.41%, predicted protein content of an external validation set of different varieties with R 2 = 0.90, root‐mean‐square error of prediction (RMSEP) = 0.42% and bias = −0.02%. However, when grains with a wider range of moisture content (7.50%–17.75%) were used for validation, prediction errors increased with R 2 = 0.72 RMSEP = 0.87% and bias = −0.32%. Inclusion of grains with a wider moisture range to the calibration set improved the performance of the calibration model with a R 2 = 0.83, RMSEP = 0.67% with a bias of −0.04%. Conclusions Variation of moisture content in grains affected the performance of the NIR protein calibration model. Likewise, inclusion of moisture variation in the calibration sample set improved the robustness of the model. Significance and novelty In addition to the traits of interest, variation of other physical or chemical traits should also be considered for inclusion into the calibration sample set to improve the robustness of NIR calibration models. This is especially important when NIR spectroscopy methods are developed for evaluation of breeding populations as the future grain samples from numerous crosses may be substantially diverse.