To accurately, safely, and efficiently evaluate the bitterness of Traditional Chinese Medicines (TCMs), a robust predictor was developed using robust partial least squares (RPLS) regression method based on data obtained from an electronic tongue (e-tongue) system. The data quality was verified by the Grubb’s test. Moreover, potential outliers were detected based on both the standardized residual and score distance calculated for each sample. The performance of RPLS on the dataset before and after outlier detection was compared to other state-of-the-art methods including multivariate linear regression, least squares support vector machine, and the plain partial least squares regression. Both R2 and root-mean-squares error (RMSE) of cross-validation (CV) were recorded for each model. With four latent variables, a robust RMSECV value of 0.3916 with bitterness values ranging from 0.63 to 4.78 were obtained for the RPLS model that was constructed based on the dataset including outliers. Meanwhile, the RMSECV, which was calculated using the models constructed by other methods, was larger than that of the RPLS model. After six outliers were excluded, the performance of all benchmark methods markedly improved, but the difference between the RPLS model constructed before and after outlier exclusion was negligible. In conclusion, the bitterness of TCM decoctions can be accurately evaluated with the RPLS model constructed using e-tongue data.

Tools to define the active ingredients and flavors of Traditional Chinese Medicines (TCMs) are limited by long analysis times, complex sample preparation and a lack of multiplexed analysis. The aim of the present study was to optimize and validate an electronic tongue (E-tongue) methodology to analyze the bitterness of TCMs. To test the protocol, 35 different TCM concoctions were measured using an E-tongue, and seven replicate measurements of each sample were taken to evaluate reproducibility and precision. E-tongue sensor information was identified and classified using analysis approaches including least squares support vector machine (LS-SVM), support vector machine (SVM), discriminant analysis (DA) and partial least squares (PLS). A benefit of this analytical protocol was that the analysis of a single sample took <15 min for all seven sensors. The results identified that the LS-SVM approach provided the best bitterness classification accuracy (binary classification accuracy, 100%; ternary classification accuracy, 89.66%). The E-tongue protocol developed showed good reproducibility and high precision within a 6 h measurement cycle. To the best of our knowledge, this is the first study of an E-tongue being applied to assay the bitterness of TCMs. This approach could be applied in the classification of the taste of TCMs, and serve important roles in other fields, including foods and beverages.

This paper focuses on determining the authenticity and identifying the species of Fritillariae cirrhosae using electronic nose, electronic tongue, and electronic eye sensors, near infrared and mid-level data fusion. 80 batches of Fritillariae cirrhosae and its counterfeits (including several batches of Fritillaria unibracteata Hsiao et K.C. Hsia, Fritillaria przewalskii Maxim, Fritillaria delavayi Franch and Fritillaria ussuriensis Maxim) were initially identified by Chinese medicine specialists and by criteria in the 2020 edition of Chinese Pharmacopoeia. After obtaining the information from several sensors we constructed single-source PLS-DA models for authenticity identification and single-source PCA-DA models for species identification. We selected variables of interest by VIP value and Wilk’s lambda value, and we subsequently constructed the three-source fusion model of intelligent senses and the four-source fusion model of intelligent senses and near-infrared spectroscopy. We then explained and analyzed the four-source fusion models based on the sensitive substances detected by key sensors. The accuracies of single-source authenticity PLS-DA identification models based on electronic nose, electronic eye, electronic tongue sensors and near-infrared were respectively 96.25%, 91.25%, 97.50% and 97.50%. The accuracies of single-source PCA-DA species identification models were respectively 85%, 71.25%, 97.50% and 97.50%. After three-source data fusion, the accuracy of the authenticity identification of the PLS-DA identification model was 97.50% and the accuracy of the species identification of the PCA-DA model was 95%. After four-source data fusion, the accuracy of the authenticity of the PLS-DA identification model was 98.75% and the accuracy of the species identification of the PCA-DA model was 97.50%. In terms of authenticity identification, four-source data fusion can improve the performance of the model, while for the identification of the species the four-source data fusion failed to optimize the performance of the model. We conclude that electronic nose, electronic tongue, electronic eye data and near-infrared spectroscopy combined with data fusion and chemometrics methods can identify the authenticity and determine the species of Fritillariae cirrhosae. Our model explanation and analysis can help other researchers identify key quality factors for sample identification. This study aims to provide a reference method for the quality evaluation of Chinese herbs.

No abstract