Ginsenoside Rf, a tetracyclic triterpenoid only present in Panax ginseng, has been proven to relieve lipid metabolism and inflammatory reactions, which can be a potential treatment for nonalcoholic fatty liver disease (NAFLD). Therefore, this study aimed to reveal the underlying mechanisms of ginsenoside Rf in the treatment of early-stage NAFLD (NAFL) by using a bioinformatics method and biological experiments. Target genes associated with NAFL were screened from the Gene Expression Omnibus (GEO) database, a database repository of high-throughput gene expression data and hybridization arrays, chips, and microarrays. Subsequently, gene set enrichment analysis was performed by using Gene Ontology enrichment analysis tool. Then, the binding capacity between ginsenoside Rf and NAFL-related targets was evaluated by molecular docking. Finally, the FFA-induced HepG2 cell model treated with ginsenoside Rf was adopted to verify the effect of ginsenoside Rf and the related mechanisms. There were 41 common differentially expressed genes in the GEO dataset. Gene Ontology and Reactome pathway enrichment analysis of the differentially expressed genes showed that many pathways could be related to the pathogenesis of NAFL, including those participating in the cytokine-mediated signaling pathway, G protein-coupled receptor signaling pathway, and response to lipopolysaccharide. Finally, the qRT-PCR analysis results indicated that ginsenoside Rf therapy could ameliorate the transcription of ANXA2, BAZ1A, DNMT3L and MMP9. Our research discovered the relevant mechanisms and basic pharmacological effects of ginsenoside Rf in the treatment of NAFL. These results might facilitate the development of ginsenoside Rf as an alternative medication for NAFL.

Many plants have multiple chemical components and multiple targets, and their potential effects on diseases are the integrative effects of multiple targets. How to systematically reveal the integrated multi‐targets effect of plants on diseases is not only a challenge, but also an innovation. This study developed a novel research method based on artificial intelligence and took hawthorn as an example; a deep auto‐encoding neural network model was used to encode the expression levels of multiple common targets between hawthorn and atherosclerosis in each cell of the single‐cell transcriptome of atherosclerotic perivascular adipose tissue (PVAT) as an integrated value (MTIS). The landscape and quantitative mapping of multi‐targets potential integrated effect of plants on disease at the single‐cell level would be achieved based on this innovative approach, and in‐depth analysis such as MTIS comparisons, MTIS‐pseudotime difference analysis, cell communication analysis, and immune infiltration analysis, was performed to reveal the potential mechanism and landscapes of hawthorn on the PVAT microenvironment of atherosclerotic. Due to many plants for disease having multiple chemical compositions and multiple targets, the novel method proposed in this study may have a wide range of applications.