ABSTRACT
Packaged coconut water offers various options, from pure to those with added sugars and other additives. While the purity of coconut water is esteemed for its health benefits, its popularity also exposes it to potential adulteration and misrepresentation. To address this concern, our study combines Fourier transform infrared spectroscopy (FTIR) and machine learning techniques to detect potential adulterants in coconut water through classification models. The dataset comprises infrared spectra from coconut water samples spiked with 15 different types of potential sugar substitutes, including: sugars, artificial sweeteners, and sugar alcohols. The interaction of infrared light with molecular bonds generates unique molecular fingerprints, forming the basis of our analysis. Departing from previous research predominantly reliant on linear-based chemometrics for adulterant detection, our study explored linear, non-linear, and combined feature extraction models. By developing an interactive application utilizing principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE), non-targeted sugar adulterant detection was streamlined through enhanced visualization and pattern recognition. Targeted analysis using ensemble learning random forest (RF) and deep learning 1-dimensional convolutional neural network (1D CNN) achieved higher classification accuracies (95% and 96%, respectively) compared to sparse partial least squares discriminant analysis (sPLS-DA) at 77% and support vector machine (SVM) at 88% on the same dataset. The CNN's demonstrated classification accuracy is complemented by exceptional efficiency through its ability to train and test on raw data.
Subject(s)
Cocos , Deep Learning , Sugars , Sweetening Agents , Spectroscopy, Fourier Transform Infrared/methods , Sweetening Agents/analysis , Cocos/chemistry , Sugars/analysis , Sugar Alcohols/analysis , Principal Component Analysis , Food Contamination/analysis , Machine Learning , Water/chemistry , Water/analysisABSTRACT
A single-laboratory validation study was conducted for the LC post-column oxidation analysis of paralytic shellfish toxins (PST): saxitoxin (STX); neosaxitoxin (NEO); gonyautoxins (GTX) 1-5; decarbamoyl gonyautoxins (dcGTX) 2 and 3; decarbamoyl saxitoxin (dcSTX); and N-sulfocarbamoyl-gonyautoxin-2 and 3 (C1 and C2) in mussels (Mytilus edulis), soft shell clams (Mya arenaria), scallops (Placopectin magellanicus), and oysters (Crassostrea virginicus). The instrumental technique was developed for the analysis of PST in shellfish as an alternative to the precolumn oxidation method, AOAC Official Method 2005.06, and a replacement for the current AOAC biological method 959.08. The method used reversed-phase LC with post-column oxidation and fluorescence detection. Test materials for method recovery were prepared by fortification of blank material with a cocktail of PST. Materials used to determine method repeatability and intermediate precision were prepared by blending blank material with naturally incurred material. The target total toxicity levels evaluated in the study were 0.40, 0.80, and 1.60 mg STX x diHCl equivalents per kilogram [(eq/kg) 1%, 1, and 2 times the regulatory limit]. Linearity, recovery, and within-laboratory precision parameters of the method were evaluated. Correlation coefficients of the calibration curves for all toxins studied were > 0.99. Total toxin recovery ranged from 94 to 106% at the three levels of interest. Repeatability and intermediate precision RSD ranged from 2 to 7% and 2 to 8%, respectively. The method LOD and LOQ (assuming the presence of all toxins) were determined to be equivalent to 0.18 and 0.39 mg STX x diHCl eq/kg. The method is intended for a regulatory framework and will be submitted for an AOAC collaborative study.