¹H NMR spectroscopy with one-class and discriminant chemometric models for detecting sugar adulteration in coconut...

Sugars may be added to coconut water to enhance sweetness or mask product dilution, necessitating the development of reliable methods for detecting such adulteration. This study evaluated proton nuclear magnetic resonance (¹H NMR) spectroscopy combined with chemometric approaches to detect coconut water adulterated with high-fructose corn syrup (HFCS), cane syrup, glucose, fructose, and sucrose. A total of 120 authentic fresh coconut water samples and 150 adulterated samples (5-50% v/v, using sugar solutions adjusted to 5.1°Brix) were analysed. The one-class approach using data-driven soft independent modelling of class analogy (DD-SIMCA) showed strong capability to identify samples belonging to the authentic class, with 23 out of 24 authentic samples correctly identified (sensitivity = 0.958). However, specificity for rejecting adulterated samples was lower (0.700), with 45 out of 150 adulterated samples incorrectly accepted as authentic, resulting in an overall efficiency of 0.819. For binary discrimination, support vector machine (SVM) outperformed partial least squares-discriminant analysis (PLS-DA), misclassifying only three samples in the prediction set (balanced accuracy = 0.946) compared with ten misclassifications for PLS-DA (balanced accuracy = 0.817). For multi-class discrimination of adulterated samples, both models performed well, with PLS-DA correctly classifying all samples (balanced accuracy = 1.00), likely due to distinct spectral profiles among sugar types. Overall, the results demonstrate that ¹H NMR spectroscopy combined with chemometric approaches provides a rapid and promising method for detecting undeclared sugar addition in coconut water.