Quality evaluation of the Salvia miltiorrhiza wine-frying process based on electronic eye, electronic nose, and untargeted metabolomics combined with multivariate fusion algorithm

Wine-frying is a traditional processing method for Salvia miltiorrhiza ( S. miltiorrhiza ), but the dynamic quality changes during the process lack objective evaluation methods. This study integrates electronic eye, electronic nose, and untargeted metabolomics to analyze the color, odor, and metabolite changes of S. miltiorrhiza during frying. Results showed that, as frying time increases, the color parameter a * decreases, while b * and c * first decrease and then increase. The electronic nose tentatively annotated 20 volatile compounds, with (E)-2-penten-1-ol and acetoin showing significant increases. The orthogonal partial least squares discriminant analysis based on odor successfully distinguished different frying degrees. Untargeted metabolomics annotated 1486 metabolites, among which dihydromacarpine and saponarin were screened as processing-sensitive differential metabolites showing increasing trends with frying time. Volcano plot revealed that frying for 8–12 min was a critical stage for significant metabolite changes. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicated that frying influenced pathways such as isoflavonoid biosynthesis. Correlation analysis established significant relationships between color, odor, and metabolites. A preliminary random forest-based discrimination model was constructed using multi-source fusion variables and optimized by genetic and particle swarm optimization algorithms. The model showed promising internal discrimination performance in the current limited dataset. This study provides a preliminary multi-dimensional framework for evaluating the wine-frying process of S. miltiorrhiza , which requires further validation using larger and more diverse sample sets.