Prediction of total protein and intact casein in cheddar cheese using a low-cost handheld short-wave near-infrared spectrometer

This study aims to develop and validate rapid quantification technique for intact casein (IC) and total protein in cheddar cheese using SCiO (TM), a low-cost short-wave near-infrared spectrometer. SCiO (TM) was used to collect the spectral scans (from 740 nm to 1070 nm) of cheddar cheese (N = 49). Reference values of IC and total protein were quantified by the Kjeldahl method. The spectra acquired from the SCiO (TM) were preprocessed using standard normal variate and second derivation, and subsequently partial least squares regression (PLSR) based models were developed (n(cal) = 35). In addition, wavelength selection methods (backward elimination, multiple linear regression-forward selection, and inteval PLSR) were applied to further improve the model performances. Based on external validation (n(val) = 14), IC prediction resulted in a range of root mean square error of prediction (RMSEP) between 0.91 and 1.58 g/100 g cheese, and residual prediction deviation (RPD) between 1.2 and 2.1. For total protein prediction, the RMSEP ranged from 0.62 to 0.88 g/100 g cheese, while the RPD was found between 1.4 and 2.0. The prediction results indicated that SCiO (TM) can provide useful quantification tool for a rapid prediction of IC and total protein in cheddar cheese.