Background
Evaluating traditional Chinese medicine (TCM) quality is a powerful method to ensure TCM safety. TCM quality evaluation
Conclusions
In the present study, traditional sensory characterization and modern standardized processes based on production process and pharmacological efficacy evaluation were integrated for use in the assessment of TCM quality. Multidimensional quality evaluation indices were integrated with a machine learning model to identify key quality indices and their corresponding weight coefficients, to establish a multiweighted multi-index and comprehensive quality index, and to construct a QCEI-based model for grading TCM quality. Our results could facilitate and guide the development of TCM quality control research.
Methods
The area of origin, number of growth years, and harvest season are considered key TCM quality attributes. In this study, licorice was the model TCM to investigate the quality indicators associated with key factors that are considered to influence TCM quality using multivariate statistical analysis, identify biological-evaluation-based pharmacological activity indicators by network pharmacology, establish real quality indicators, and develop a QCEI-based model for grading TCM quality using a machine learning model. Finally, to determine whether different licorice quality grades differently reduced the inflammatory response, TNF-α and IL-1β levels were measured in RAW 264.7 cells using ELISA analysis.
Results
The 21 quality indices are suitable candidates for establishing a method for grading licorice quality. A computer model was established using SVM analysis to predict the TCM quality composite evaluation index (TCM QCEI). The tenfold cross validation accuracy was 90.26%. Licorice diameter; total flavonoid content; similarities of HPLC chromatogram fingerprints recorded at 250 and 330 nm; contents of liquiritin apioside, liquiritin, glycyrrhizic acid, and liquiritigenin; and pharmacological activity quality index were identified as the key indices for constructing the model for evaluating licorice quality and determining which model contribution rates were proportionally weighted in the model. The ELISA analysis results preliminarily suggest that the inflammatory responses were likely better reduced by premium-grade than by first-class licorice. Conclusions: In the present study, traditional sensory characterization and modern standardized processes based on production process and pharmacological efficacy evaluation were integrated for use in the assessment of TCM quality. Multidimensional quality evaluation indices were integrated with a machine learning model to identify key quality indices and their corresponding weight coefficients, to establish a multiweighted multi-index and comprehensive quality index, and to construct a QCEI-based model for grading TCM quality. Our results could facilitate and guide the development of TCM quality control research.