The varying quality of Paris polyphylla var. yunnanensis (PPY) growth years available on the market significantly impedes its medicinal efficacy. To identify and evaluate PPY from different years, this study established a multi-dimensional spectroscopy data fusion combined with metabolomics. Firstly, multi-dimensional spectral images were created using MIR, NIR and data fusion techniques. Then, partial least squares discriminant analysis (PLS-DA), support vector machines (SVM) and deep learning ResNet model, were applied to classify PPY based on a multi-dimensional spectral data fusion. Finaly, metabolomic analysis of saccharides and flavonoids was used for evaluating and screening the markers of different growth years. The results indicated that preprocessing spectra outperformed raw spectra in the PLS-DA and SVM, and the combining multiple preprocessing techniques yielded better results than single method. Data fusion of NIR and MIR spectra achieved 100% accuracy in both training and test for the PLS-DA and SVM models. Moreover, 2DCOS and 3DCOS synchronous spectra, coupled with ResNet without preprocessing, enabled high classification accuracy for PPY from different growth years. In the metabolomics analysis, eleven markers — comprising five saccharides and six flavonoids, which were screened for distinguishing PPY, with their content varying across different growth years. This study provides a reference for the rapid identification and evaluation of Chinese herbal medicines. • A multi-dimensional near-infrared and mid-infrared spectroscopy data fusion were generated. • Machine learning was applied to identify Paris polyphylla var. yunnanensis from different growth years. • 2DCOS and 3DCOS synchronous spectra coupled to ResNet could achieve high classification accuracy. • Five saccharides and six flavonoids were screened for distinguishing Paris polyphylla var. yunnanensis .
Yang et al. (Fri,) studied this question.