Qualitative and quantitative analysis of Paris polyphylla var. yunnanensis from different origin using...

Geographical traceability is crucial to ensuring consistent quality of Paris polyphylla var. yunnanensis (PPY) from different regions. In this study, a multi-dimension FT-MIR and NIR spectroscopy, data fusion combined with machine learning were used for identifying PPY qualitatively and quantitatively. A multi-dimensional spectrum includes 1D, 2DCOS, 3DCOS (synchronous, asynchronous and integrative) were generated. FT-MIR, NIR and data fusion coupled to traditional machine learning includes principal component analysis (PCA), t-distributed Stochastic Neighbor Embedding (t-SNE) partial least squares discriminant analysis (PLS-DA), orthogonal partial least squares discriminant analysis (OPLS-DA) were used to identify PPY from different regions. Meanwhile, deep learning model ResNet based on multi-dimensional spectral images were established for different maps. FT-MIR, NIR and data fusion were applied to determine total saponins. The results shows that PCA and t-SNE can reveal the relationship of the general chemical information of PPY. PLS-DA and OPLS-DA based on data fusion had the best performance with greater R²X, R²Y and Q², and smaller RMSEE, RMSECV and RMSEP, and higher accuracy of training and test set. Multi-dimensional spectroscopy with no processing had good performance in classification. Both synchronous 2DCOS and 3DCOS images for FT-MIR and NIR could effectively classify samples. PLSR based on data fusion coupled to the combination of pretreatment spectral multiple scattering correction and first derivative had the better predictive results than others, achieving an RPD of 1.98, with Rc² and Rp² of 0.92 and 0.78, respectively. This study provides a rapid and robust strategy for quality control of Traditional Chinese medicine.