Single-cell and bulk RNA-seq with machine learning reveals CCNB1/CDK1-mediated G2/M arrest as key anti-HCC mechanism of Fuzheng Xiaozheng Prescription.

BACKGROUND: Hepatocellular carcinoma (HCC) remains a leading cause of cancer-related mortality worldwide, with a 5-year survival rate below 20 % for advanced-stage patients. Fuzheng Xiaozheng Prescription (FZXZP), a traditional Chinese medicine compound, has demonstrated promising anti-HCC activity in preliminary studies, but its key bioactive components and molecular targets remain undefined. PURPOSE: To systematically elucidate the active constituents and mechanistic targets of FZXZP against HCC using an integrated bulk RNA, single cell RNA sequencing (scRNA-seq), network pharmacology and experimental validation approach. METHODS: The anti-tumor efficacy of FZXZP was evaluated in a murine HCC model and in human HCC cell lines. Untargeted metabolomics (LC-MS/MS) was performed to identify bioactive components in FZXZP aqueous extract and drug-containing serum. Network pharmacology analyses were then conducted to predict potential molecular targets associated with these bioactive constituents. Transcriptomic analyses integrated bulk RNA-seq data from TCGA-LIHC and GEO (GSE87630, GSE149614) with scRNA-seq data (GSE149614). Differential expression analysis, weighted gene co-expression network analysis, univariate Cox regression and network pharmacology were integrated to identify candidate genes. Three machine-learning algorithms (XGBoost, maximal clique centrality, and Eccentricity) were used to refine key genes. Key findings were rigorously validated in independent cohorts at the single-cell level, and confirmed through in vitro and in vivo experiments, including flow cytometry, Western blot, and quantitative real-time PCR (qRT-PCR). Loss- and gain-of-function experiments were performed via cell transfection to validate key targets and mechanisms. Finally, molecular docking, molecular dynamics (MD) simulations and surface plasmon resonance (SPR) assay assessed compound-target binding. RESULTS: FZXZP treatment markedly inhibited HCC progression in murine models, significantly improving liver histopathology and serum ALT/AST levels. In vitro, FZXZP effectively suppressed HCC cell proliferation, migration, and invasion. A total of 134 bioactive components were identified in FZXZP aqueous extract and drug-containing serum, corresponding to 1181 predicted targets. Integration of 1208 differentially expressed genes, 297 hub genes, 9999 risk-associated genes, and 1181 drug targets yielded 17 candidate genes, significantly enriched in cell cycle regulation signaling pathways. Machine-learning algorithms consistently selected CCNB1 as the core regulatory gene. scRNA-seq analysis demonstrated that fibrotic cells exhibited the most pronounced intercellular communication with other cell types, and CCNB1 expression was significantly elevated within the fibrotic cell population in HCC. CCNB1 showed robust diagnostic performance (AUC of 0.971 in TCGA, 0.784 in GSE87630). Mechanistically, FZXZP induced G2/M phase arrest and downregulated CCNB1 and CDK1 expression at both the mRNA and protein levels. In cell models involving gene knockdown and overexpression, CCNB1 has been validated as an oncogene and identified as a key target for FZXZP against HCC. Molecular docking, MD simulations and SPR assay further confirmed stable binding between 7-Methoxyflavone, a principal active component, and CCNB1(binding energy:-83.218 kJ/mol; KD=1.72 µM). 7-Methoxyflavone alone recapitulated the anti-HCC effects, supporting its role as a representative bioactive compound of FZXZP. CONCLUSION: FZXZP exerts its anti-HCC effects primarily through its representative component, 7-Methoxyflavone, by suppressing CCNB1/CDK1 and consequently triggering G2/M cell cycle arrest. These findings establish CCNB1 as a therapeutic target and provide a mechanistic rationale for FZXZP in HCC treatment.