SiO₂ Nanolayer Regulated Ag@Cu Core-Shell SERS Platform Integrated Machine Learning for Intelligent Identification of Jujuboside A, Saikosaponin A and Timosaponin A-III

Abstract

An effective SERS-based detection method assisted by machine learning algorithm has been developed to intelligently identify the pharmacodynamic substances in traditional Chinese medicine (TCM) which aided in the quality control and identification of TCM. In this work, the ultrasensitive grating-like core-shell Ag@SiO2@Cu@moth wing (Ag@SiO2@Cu@MW) SERS platform was explored which can precisely control the intergap distances by adjusting the thickness of the SiO2 layer. This SERS platform demonstrated reproducibility and reliability, with a low relative standard deviation (RSD) of 7.38% and an enhancement factor (EF) of 2.49 × 107. The optimized Ag30@SiO2(4)@Cu20@MW substrate were then applied in the accurately analyzing pharmacodynamic substances. Specifically, the label-free SERS analysis showed the distinct spectral features for Jujuboside A, Saikosaponin A and Timosaponin A-III. Machine learning algorithms, such as principal component analysis (PCA), decision tree (DT), support vector machine (SVM), k-nearest neighbors (kNN) were employed and further in differentiating with the three pharmacodynamic substances Raman spectrum groups. These results indicate that SERS technology in combination with machine learning algorithms can not only achieve rapid and accurate detection of different types of pharmacodynamic substances, but also promote the modernization and international application of TCM.

Keywords

• Grating-like core-shell structure
• machine learning
• pharmacodynamic substances in traditional Chinese medicine
• SERS