Abstract
The leaves of Morus alba Linn., which is also known as white mulberry, have been commonly used in many of traditional systems of medicine for centuries. In traditional Chinese medicine (TCM), mulberry leaf is mainly used for anti-diabetic purpose due to its enrichment in bioactive compounds such as alkaloids, flavonoids and polysaccharides. However, these components are variable due to the different habitats of the mulberry plant. Therefore, geographic origin is an important feature because it is closely associated with bioactive ingredient composition that further influences medicinal qualities and effects. As a low-cost and non-invasive method, surface enhanced Raman spectrometry (SERS) is able to generate the overall fingerprints of chemical compounds in medicinal plants, which holds the potential for the rapid identification of their geographic origins. In this study, we collected mulberry leaves from five representative provinces in China, namely, Anhui, Guangdong, Hebei, Henan and Jiangsu. SERS spectrometry was applied to characterize the fingerprints of both ethanol and water extracts of mulberry leaves, respectively. Through the combination of SERS spectra and machine learning algorithms, mulberry leaves were well discriminated with high accuracies in terms of their geographic origins, among which the deep learning algorithm convolutional neural network (CNN) showed the best performance. Taken together, our study established a novel method for predicting the geographic origins of mulberry leaves through the combination of SERS spectra with machine learning algorithms, which strengthened the application potential of the method in the quality evaluation, control and assurance of mulberry leaves.
Water and ethanol extracts of mulberry leaves have different SERS fingerprints.
Extracts of mulberry leaves from different geographic origins have different SERS fingerprints.
SERS spectra of mulberry leaf extracts can be clustered into different groups via OPLS-DA algorithm according to geographic origins and extraction methods.
Supervised machine learning algorithms are capable of predicting the geographic origins of mulberry leaf extracts via SERS spectra.
HIGHLIGHTS
Communicated by Ramaswamy H. Sarma
Acknowledgements
We would like to thank the anonymous reviewers for their constructive comments and suggestions that significantly improved the quality of this study.
Author’s contributions
LW, DQT and CYW conceived and designed the experiments and provided the platform and resources. LW and DQT were responsible for project administration. ZWM, JWT, QHL, JYM, RQ, and YD carried out the experimental work. LW, ZWM, JWT, and QHL performed the data analysis. LW, ZWM, JWT and QHL wrote and revised the manuscript. All authors read and approved the final manuscript.
Disclosure statement
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.