Synergizing Network Pharmacology and Pan-Cancer Analysis in TCM Repositioning for Tumor Therapy

BackgroundRepositioning Traditional Chinese Medicine (TCM) for cancer treatment, addressing heterogeneity through synergistic effects, aligns with the evolving trend of combination therapy. However, the complexity of TCM and the lack of methodology hinders the elucidation of TCM treatment mechanisms and potential indications. The research aims to construct develop a comprehensive method combining network pharmacology and pan-cancer analysis (NetPharm-PanCan) for TCM repositioning, exemplified by the Astragali Radix-Curcumae Rhizoma (ARCR) herb pair. MethodThe TCM-component-gene network was constructed using Cytoscape3.7.2 with gene screening based on components from TCMSP and targets predicted via the Swiss Target Prediction database. The core gene set of the ARCR (CGSARCR) was identified by analyzing network via the dual algorithm, namely Degree and MCODE followed by GO and KEGG enrichment analyses. The pan-cancer analysis encompassing Gene Set Variation Analysis (GSVA), immune infiltration correlation, cancer pathway analysis, and Gene Set Enrichment Analysis (GSEA) was unveiled to identify the potential indications. Multivariate cox regression analysis was employed for identifying prognostic genes, followed by the modeling for potential therapeutic indications using the R software. The Metescape database was harnessed to reveal similarities and differences in the mechanisms for prognostic genes associated with potential indications. The XGBoost algorithm was used to construct an indication prediction system according to the analysis results above. ResultThe CGSARCR comprises 28 genes targeting all ARCR components. The pan-cancer analysis revealed that the CGSARCR showed significantly higher tumor scores than normal tissue scores across nine different types of cancer, including THCA, KIRC, LUAD, COAD, BRCA, STAD, ESCA, and others. The CGSARCR correlated strongly with cancer-related pathways and the immune microenvironment. Prognostic models evaluate the potential of these indications as follows: THCA> KIRC> LUAD> COAD> BRCA> STAD> ESCA, with enrichment analysis suggesting KIRC and LUAD as the most potential indications of ARCR. The XGBoost-based system achieved high predictive accuracy (training AUC: 0.985, testing AUC: 0.96, training logloss: 0.21, testing logloss: 0.25). ConclusionThe NetPharm-PanCan method provides a robust, network-based pan-cancer analysis framework for TCM repositioning in cancer research. It provides theoretical foundations and practical tools for TCM-based drug development, with potential applicability to broader drug repurposing efforts.