Exploring the mechanism of epigallocatechin-3-gallate on ovarian cancer based on network pharmacology and experimental validation

Background Ovarian cancer is associated with a poor prognosis, largely due to late-stage diagnosis and the development of therapy resistance. The green tea polyphenol epigallocatechin-3-gallate (EGCG) shows anti-tumor potential, but its targets and mechanisms in ovarian cancer remain unclear. This study aims to identify the key molecular targets and elucidate the underlying mechanisms of EGCG’s action in ovarian cancer. Methods Differentially expressed genes (DEGs) in the GSE18520 dataset were identified using the limma package (|log 2 fold change| > 2, adjusted p < 0.05). Weighted gene co-expression network analysis (WGCNA) was performed with key parameters: soft thresholding power β = 12 (scale-free R 2 = 0.801), unsigned network type, deepSplit = 2, minModuleSize = 15, and mergeCutHeight = 0.25. Hub genes were screened by intersecting DEGs, WGCNA brown module genes (|Module Membership| > 0.8, |Gene Significance| > 0.2), and EGCG targets from the Comparative Toxicogenomics. Three machine learning algorithms, including least absolute shrinkage and selection operator (LASSO) (5-fold cross-validation), support vector machines-recursive feature elimination (SVM-RFE), and random forest (RF) (Gini index), were applied to identify core targets. Cells were randomly assigned to Control, EGCG-treated, and EGCG + potassium voltage-gated channel subfamily H member 2 (KCNH2)-overexpression groups. In vitro experiments (Cell Counting Kit-8 (CCK-8), colony formation, flow cytometry) and molecular docking (AutoDock Vina) were used for validation. Results A total of 1204 DEGs were identified (|log 2 fold change| > 2, adjusted p < 0.05). WGCNA identified 8 co-expression modules, among which the brown module (containing 386 genes) showed the strongest correlation with ovarian cancer. Intersection analysis yielded 19 common genes, and machine learning further narrowed this to KCNH2 as the sole overlapping target. KCNH2 was up-regulated in ovarian cancer tissues and cell lines ( p < 0.01). Molecular docking predicted a potentially favorable interaction between EGCG and KCNH2, with a binding energy of −6.7 kcal/mol. Functionally, EGCG inhibited ovarian cancer cell viability ( p < 0.01), and KCNH2 overexpression reversed EGCG-induced decrease in KCNH2 expression ( p < 0.05). Additionally, EGCG reduced colony formation and induced apoptosis ( p < 0.001), while KCNH2 overexpression reversed these effects ( p < 0.01). Conclusion Based on bioinformatic and in vitro analyses, EGCG appears to suppress ovarian cancer cell growth and promote apoptosis partly by targeting and down-regulating KCNH2. These preliminary findings propose the EGCG-KCNH2 axis as a candidate mechanism warranting further preclinical investigation. However, given that the current evidence is limited to computational predictions and cell-based assays, additional studies in animal models and clinical specimens are necessary before any therapeutic implications can be considered. Nevertheless, this work provides a testable mechanistic hypothesis for EGCG’s anti-ovarian cancer activity, distinguishing it from non-specific anti-proliferative effects.