Exploring the feasibility of multi-elements coupled with chemometrics for discriminating the geographical origins of oysters (Crassostrea ariakensis)

This study explored the efficacy of multi-elements combined with chemometrics to discriminate the geographical origins of oysters (Crassostrea ariakensi). We determined 52 elements in 166 samples from four regions along the southeast coast of China. Significant regional variations of 51 elements were revealed (P < 0.05), while the principal component analysis (PCA) provided no clear regional delineations. The training models (n = 117) established on linear discriminant analysis (LDA), partial least square discriminant analysis (PLS-DA), and random forest (RF) uniformly achieved 100% predictive accuracy. The cross-validation accuracies of the final models (n = 166) derived from LDA, PLS-DA, and RF were 100%, 100%, and 99.4%, respectively. Even with the models simplified to 8 elements (Zn, Al, K, Cd, Cu, Rb, B, and Ag), high predictive and cross-validation accuracies were maintained, underscoring the robustness and algorithm flexibility of elemental profiling for accurately identifying the geographical origins of oysters.