Prediction of crude protein and oil content of soybeans using Raman spectroscopy

Abstract While conventional chemical analysis methods for food nutrients require time-consuming, labor-intensive, and invasive pretreatment procedures, Raman spectroscopy can be used to measure a variety of food components rapidly and non-destructively without supervision from experts once the instrument has been calibrated. The purpose of this study was to develop an optimal prediction model for determining the protein and oil contents of soybeans using a dispersive Raman spectroscopy method. In general, the crude oil content of soybeans is chemically determined using the Soxhlet extraction method, while the semimicro-Kjeldahl method and an auto protein analyzer have been used to assess crude protein content. In the present study, Raman spectra were measured in the 200–1800 cm−1 wavenumber range and partial least squares (PLS) analysis methods were used to develop optimal models for predicting the crude protein and oil contents of soybeans. The resultant PLS models that used the effective wavenumber regions determined by intermediate PLS (iPLS) method were better than those models developed using the entire wavenumber range under investigation. The R p 2 and SEP of the optimal PLS model for crude protein content were 0.916 and 0.636%, respectively. Likewise, the R p 2 and SEP for crude oil content were 0.872 and 0.759%, respectively. The result suggests that the conventional Raman techniques investigated in this study can be applied to the prediction of soybean crude protein and oil content.