Accumulation mechanism of metabolites markers identified by machine learning between Qingyuan and Xiushui counties in Polygonatum cyrtonema Hua

Abstract

Abstract Polygonatum cyrtonema Hua (P. cyrtonema) is well known for its high medicinal value due to a wide variety of secondary metabolites. Nonetheless, the unclearness persists regarding the distribution and buildup processes of these metabolites across various regions. Using UPLC-ESI-MS/MS, a grand total of 482 metabolites were detected and identified in this research. Cluster analysis showed distinct metabolite profiles between Qingyuan County and Xiushui County. The identification of secondary metabolites, such as flavonoids, phenolic acids, and lignans, between the two regions was performed using support vector machine-recursive feature elimination (SVM-RFE) and random forest (RF) machine learning techniques. Subsequently, the process of comparative transcriptomics and weighted gene co-expression analysis (WGCNA) uncovered genes associated with flavonoids such as CHI, UGT1, UGT2, ERF, as well as phenylpropane-related genes UGT3 and NAC. In addition, by comparing transcriptomes, four genes (PcOMT10/11/12/13) were selected as differentially expressed. Their impact on metabolic fluxes of the phenolpropane pathway was confirmed using a transient expression system in tobacco. The findings enhanced our comprehension of the variation in accumulation of secondary metabolites mediated by phenylpropanoids across various locations, and offered valuable genetic assets for the synthesis of bioactive compounds.