A Multi-omics approach combining GC-MS, LC-MS, and FT-NIR with chemometrics and machine learning for metabolites systematic profiling and geographical origin tracing of Artemisia argyi Folium

Artemisia argyi Folium (AAF) is a widely used functional aromatic medicinal plant with significant health, culinary, and therapeutic value. However, geographical variations in environmental factors notably affect its metabolite production, making comprehensive analysis of AAF quality essential. This study aims to systematically characterize AAF from different geographical regions in China using a multi-omics analysis platform. By employing chemometrics and machine learning, we identify differentiated compounds in AAF and develop rapid, non-destructive models to determine its geographic origin. Comprehensive characterization of AAF samples from seven provinces in China was conducted using HS-SPME-GC-MS, UHPLC-Q-TOF-MS, and FT-NIR. Potential quality markers for volatile and non-volatile compounds were identified through mass spectrometry and chemometrics. Utilizing robust machine learning techniques, XGBoost and Stacking models were constructed for geographic tracing of AAF based on spectral data. Results revealed the identification of 66 volatile and 32 non-volatile compounds via HS-SPME-GC-MS and UHPLC-Q-TOF-MS. Chemometric analysis, including PCA and PLS-DA, highlighted 12 volatile and 12 non-volatile compounds as key quality markers. The combination of FT-NIR with machine learning demonstrated distinct spectral bands for AAF samples from different origins, achieving an accuracy of 94 % through the comparison of 130 spectral preprocessing techniques and three feature selection methods. This study indicates significant metabolomic differences in AAF based on geographic origin, emphasizing the potential of multidimensional metabolomics and machine learning in natural product quality assessment, thereby supporting germplasm development, marketing, and quality control of AAF.