A field trials-based authentication study of conventionally and organically grown Chinese yams using light stable isotopes and multi-elemental analysis combined with machine learning algorithms.

In this study, stable isotopes and multi-element signatures combined with chemometrics were used to distinguish conventional and organic Chinese yams based on field trials. Four light stable isotopes δD, δ13C, δ15N, δ18O, and 20 elements (e.g. Li, Na, Mn) were determined, then evaluated using significance analysis and correlation analysis, and modeling of various chemometrics methods. Consequently, the RandomForest model showed the best performance with AUC value of 0.972 and predictive accuracy of 97.3%, and Mn, Cr, Se, Na, δD, As, and δ15N were screened as significant variables. Moreover, many chemical components and antioxidant activity of yam samples were determined spectrophotometrically. The results indicated that organic yams had advantages in secondary metabolites such as polyphenol, flavonoid and saponin; conversely, conventional samples had more primary metabolites like protein and amino acids. Above all, this work provides a beneficial case in the authentication and quality evaluation of conventional and organic yams.