Multidimensional quality evaluation and traceability study of Fritillariae Cirrhosae Bulbus from different sources

The quality of "Fritillariae Cirrhosae Bulbus (FCB)" is influenced by its geographical origin and cultivation management. Characterizing quality differences among FCB from different sources through multidimensional analysis and establishing an accurate traceability model represent critical approaches to ensure FCB medicinal material quality. This study integrated untargeted metabolomics, alkaloid quantification, mineral nutritional element analysis, and hyperspectral imaging features to systematically reveal metabolic and compositional variations in FCB from different sources, while constructing a deep learning-based traceability model. Untargeted analysis identified significant differences in metabolite levels across FCB sources, with Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis revealing that these differential metabolites were primarily enriched in 23 pathways. Targeted alkaloid quantification demonstrated that field-collected wild specimens from Seka township (designated SK-FC) accumulated higher levels of peimisine, imperialine, and peiminine, whereas tissue-cultured regenerants from Bamei town (designated BM-TC) exhibited elevated peimine content, indicating that geographical environments and cultivation practices regulate alkaloid biosynthesis. Mineral nutritional element analysis showed that BM-TC samples had the highest elemental accumulation, likely linked to nutrient-rich culture media, while field-collected wild specimens from Chuanzhusi town (designated CZS-FC) and Anhong township artificial cultivated accessions (designated AH-AC) preferentially accumulated Al/Fe/Mn/Na and K/Mg/Zn/Cu, respectively. Most elements showed positive correlations with peiminine and peimine levels but negative correlations with peimisine and imperialine. The Residual Network (ResNet) deep learning model, constructed using hyperspectral-derived three-dimensional correlation spectroscopy (3DCOS) images, achieved 100% testing/validation accuracy and 86.67% external validation accuracy, outperforming traditional partial least squares discriminant analysis (PLS-DA) models in traceability efficacy and providing an efficient method for precise origin identification of FCB. This research establishes theoretical foundations for multidimensional quality evaluation and traceability of FCB, offering fundamental support for further development and utilization of FCB resources.