The combination of near-infrared spectroscopy (NIR) and chemometrics for qualitative and quantitative detection of additives in soy sauce

Abstract

Some producers add monosodium glutamate (MSG), yeast extract (YE), enzymatic hydrolysate of wheat (EW), and enzymatic hydrolysate of corn (EC) in traditional soy sauce (TSS) to enhance its competitiveness, seriously damaging consumer's rights. Currently, no reliable and rapid methods can detect the additives in TSS. This study successfully developed a method by combining NIR with chemometrics to solve the problem. Spectra of 30 TSS and 320 additive-added samples were analyzed by DD-SIMCA, random forest (RF), Support Vector Machine (SVM), k-Nearest Neighbor (KNN) and backpropagation artificial neural network (BP-ANN) models for qualitative classification. A competitive adaptive reweighted sampling (CARS)-optimized partial least squares (PLS) model was employed for quantitative prediction. DD-SIMCA, SVM, KNN, RF and BP-ANN achieved above 95%, classification accuracy. CARS-optimized PLS model showed strong predictability with correlation coefficient of the prediction set (Rp) above 0.98 and residual predictive deviation (RPD) above 3.0, demonstrating a reliable and rapid approach was established.

Keywords: Additive; Chemometrics; Near-infrared spectroscopy; Soy sauce.