Application of chemometrics to prediction of some wheat quality factors by near-infrared spectroscopy

Background and objectives Physical quality parameters of wheat include kernel texture and test weight, which affect classification, grading, and price. Physicochemical factors are associated with flour functionality. The objective of the work was to determine the effectiveness with which these factors could be predicted in early generations using near-infrared spectroscopy (NIRS). Wheat breeders need to know how their new genetic lines carry these factors, which carry no identifiable absorbers in the NIR region, can be predicted in early generations, by experts in the use of NIRS and its associated chemometrics. Findings With the exception of protein content, for which absorbers are plentiful, and was included to verify the spectral quality of the sample sets, none of the strictly physical factors could be reliably predicted with the most widely used chemometric options in the hands of experts in their use. Random forest algorithms were capable of prediction of all physicochemical factors, except Farinograph development time, with a reasonable degree of reliability. Conclusions Reliable predictions of quality factors in wheat that can be predicted by NIRS with satisfactory reliability are limited to chemical and physicochemical factors for which absorbers exist in the NIR region. Significance and novelty Chemical, physicochemical, and physical factors can be predicted with acceptable reliability by NIRS using a computerized spectrophotometer in association with Random Forest software. Farinograph development time remains a challenge for NIRS application.