Is protein "in the bag"? Estimating protein content in faba bean flour through a Ziploc bag using interactance NIRS

Protein content is considered one of the key indicators of faba bean quality. Near-infrared spectroscopy (NIRS) is commonly used for the rapid prediction of protein in this crop, with a typical standard error of prediction (SEP) of 0.3% w/w or better in faba bean flour. However, common sampling methods for NIRS, such as the spinning cup, still require a number of sample preparation steps, including sub-sampling of the flour, filling and packing the cup, obtaining the spectra, and cleaning the cup afterward. In this study, FT-NIR spectra (1000-2500 nm) were collected from 90 faba bean flour samples using an integrating sphere with a spinning cup, as well as through a polyethylene sample bag (Ziploc® bag) using an interactance sample gun. After subtracting the NIR spectra of an empty sample bag, partial least squares regression analysis was used to predict the protein contents of the faba bean flour from their NIR spectra. Sampling through the Ziploc® bag took approximately one-third of the analysis time compared to the spinning cup method, with the trade-off being a higher error in the predicted protein content (approximately 0.5% RMSE for cross-validation, compared to approximately 0.3% RMSECV for the spinning cup). In comparison, the laboratory error (SEL) of the reference method was 0.17%. A refined version of this in-bag analysis technique could potentially find use for screening purposes.