Origin traceability and quality assessment of licorice in Asia based on multidimensional fingerprinting and enhanced by deep learning.

Licorice (Glycyrrhizae Radix et Rhizoma) is renowned for its nutritional properties and exhibits quality variation due to environmental differences across geographical origins. Conventional analytical methodologies, such as various liquid chromatography and mass spectrometry techniques, often provide limited information or are expensive and complex. To address these challenges, a cost-effective approach has been developed using deep learning-enhanced multidimensional fingerprinting in this study. Samples from six Central Asian countries and six Chinese provinces were collected and analyzed. For the first time, chemometrics, machine learning and network pharmacology were combined to predict seven key quality markers, which were subsequently refined through Pearson correlation analysis. Deep learning models were then developed to accurately determine the geographical origin and assess the quality of licorice. The AI-enhanced approach effectively identifies potential geographical origins and quality differences based on chemical composition, offering broader applications in the authentication and quality assessment of other foods and herbal products.