Application of the Ginsenoside Multidimensional Information Library (GinMIL) Enables Accurate Characterization of Ginsenosides from Diverse Ginseng Products and Accelerates the Discovery of New Saponin Compounds

Accurate characterization of ginsenosides from ginseng relying on liquid chromatography-mass spectrometry (LC-MS) is challenging due to the lack of sufficient structural information. By machine learning techniques, we have established a ginsenoside multidimensional information library, namely, GinMIL, covering four dimensions of structural information of 579 ginsenosides. This work was designed to accurately characterize ginsenosides from Panax notoginseng products and to rapidly discover novel ginsenosides from Panax quinquefolius flowers by ion-mobility LC/MS profiling and efficient GinMIL matching on UNIFI. Consequently, we characterized 334/356/738/545 ginsenosides from three parts/two extracts/four single preparations/seven compound preparations of Panax notoginseng, respectively. 45/99/59/116 novel masses were discovered in four types of notoginseng products, respectively. Four novel ginsenosides, including three rare dimalonyl ginsenosides and one methylated malonyl ginsenoside, were isolated from Panax quinquefolius flowers by feat of GinMIL analysis. This work can verify the superiority of GinMIL, thus greatly enhancing the multicomponent characterization and the discovery of new compounds from functional herbs.