A flavonoid extracted from a plant long used in traditional Chinese medicine may act on a previously underappreciated oxidative stress hub in gastric cancer, according to a new study published in Food Science & Nutrition. By weaving together network pharmacology, bulk tumor transcriptomics, single-cell RNA sequencing, molecular docking, and laboratory experiments on gastric cancer cells, researchers identified NADPH oxidase 4, or NOX4, as a candidate regulatory node linking icariin—the principal prenylated flavonoid glycoside from Epimedium species—to the redox biology of one of the world’s most lethal malignancies. The work offers a systems-level map of how a multi-target natural compound might intersect with tumor progression, vascular remodeling, and immune contexture in the stomach.
Gastric cancer remains a formidable clinical challenge. Despite advances in surgery, chemotherapy, targeted agents, and immunotherapy, patients with advanced disease continue to face poor prognosis, and the disease is driven by interconnected processes including uncontrolled proliferation, invasion, metastasis, angiogenesis, immune escape, and remodeling of the tumor microenvironment. Icariin has previously shown anti-inflammatory, antioxidant, immunomodulatory, and antitumor properties across several cancer models, and earlier studies suggested it could influence malignant behaviors in gastric cancer cells. Yet because natural compounds typically engage dozens of targets simultaneously, the key molecular mediators of icariin’s effects in gastric cancer had never been systematically clarified. The new study set out to close that gap using an integrated computational and experimental framework.
The team began with classic network pharmacology. Potential icariin targets were retrieved from the TCMSP database and converted to official gene symbols via UniProt, while gastric cancer-related genes were defined as the union of differentially expressed genes and co-expression modules associated with the disease in the TCGA-STAD cohort. Differential expression analysis with limma revealed marked transcriptional differences between tumor and normal tissue, and weighted gene co-expression network analysis identified a turquoise module—enriched for cell cycle, DNA replication, p53 signaling, and viral carcinogenesis pathways—as most strongly correlated with the gastric cancer phenotype. Intersecting icariin targets with disease genes yielded 24 overlapping candidates, and univariate Cox regression singled out NOX4 as significantly associated with patient survival.
Follow-up analyses across large public datasets strengthened the case for NOX4’s clinical relevance. Pan-cancer profiling showed aberrant NOX4 expression in multiple tumor types, with marked upregulation in gastric cancer. In three independent cohorts—GSE15497, GSE84437, and GSE26253—high NOX4 expression consistently predicted poorer overall survival. Gene set enrichment analysis revealed that tumors with high NOX4 expression were enriched for progression-related programs such as ECM-receptor interaction, pathways in cancer, focal adhesion, and TGF-beta signaling, whereas low-NOX4 tumors leaned toward metabolic pathways including fatty acid degradation, pyruvate metabolism, and drug metabolism via cytochrome P450. NOX4 expression also rose with pathological grade and clinical stage, and a nomogram integrating NOX4 with clinical variables showed good calibration for predicting one-, three-, and five-year survival.
The study’s most striking single-cell insight came from an analysis of 142,053 cells drawn from 40 gastric tissue samples—29 tumors and 11 normal specimens—in dataset GSE183904. After quality control, Harmony-based batch correction, and clustering, cells resolved into 38 clusters spanning 11 major populations, including malignant epithelial cells, T and NK cells, macrophages, endothelial cells, fibroblasts, and myofibroblasts. InferCNV analysis separated epithelial cells into high- and low-copy-number-variation groups, confirming malignant heterogeneity, while CytoTRACE revealed diverse differentiation states across the microenvironment. Crucially, NOX4 expression was concentrated almost exclusively in endothelial cells, prompting the researchers to split this compartment into NOX4-positive and NOX4-negative subpopulations with partially distinct transcriptional identities.
These two endothelial states differed in ways that matter for tumor biology. NOX4-positive endothelial cells were enriched for angiogenesis, focal adhesion, integrin signaling, ECM-receptor interaction, PI3K-Akt signaling, and actin cytoskeleton regulation, along with elevated oxidative phosphorylation, glycolysis, and glutathione metabolism. NOX4-negative endothelial cells instead showed immune-related features such as antigen processing and presentation. Transcription factor analysis via SCENIC found higher ELK3, JUN, and JUNB activity in the NOX4-positive subset. CellChat-based communication mapping showed that NOX4-positive endothelial cells interacted more broadly with macrophages, fibroblasts, dendritic cells, and T/NK cells, with ligand-receptor axes involving SPP1, MDK, MIF, and ANGPTL2 linking them to both normal and malignant epithelial cells. A gene signature derived from this subset predicted poor survival across the TCGA-STAD, GSE84437, and GSE84433 cohorts, and consensus clustering based on the signature reproducibly separated patients into two subtypes, with the high-score cluster showing worse outcomes and enrichment for epithelial-mesenchymal transition, angiogenesis, IL6-JAK-STAT3, and inflammatory response pathways.
To translate the signature into a practical tool, the researchers built a five-gene prognostic model—comprising SPIRE1, PLCH1, KCNS3, SLC27A2, and GRP—using LASSO-Cox regression on genes shared across the three cohorts. The risk score stratified patients into high- and low-risk groups with significantly different survival in all cohorts, showed moderate time-dependent ROC performance for one-, three-, and five-year survival, and remained prognostic in multivariate models adjusting for clinicopathological variables. Notably, when the five model genes were excluded from the NOX4-positive endothelial signature, the recalculated score still correlated positively with the risk score, indicating that the model captures a broader transcriptional program of this endothelial state rather than riding on gene overlap alone. High-risk tumors also displayed lower immune scores, higher stromal scores, and differential expression of immune checkpoints and related molecules including LAG3, CD274 (PD-L1), CD276, and CD80.
Computational and wet-lab experiments then probed the icariin–NOX4 connection directly. Molecular docking using AutoDock Vina and Discovery Studio placed icariin within the predicted NOX4 binding pocket, with potential hydrogen bonds to Ser576 and Trp377, and a 100-nanosecond molecular dynamics simulation in GROMACS showed the complex settling into a stable conformation with limited ligand drift. In AGS gastric adenocarcinoma cells, icariin reduced viability in a concentration-dependent manner with an IC50 of 124.7 micromolar. Quantitative PCR and western blotting showed dose-associated reductions in NOX4 mRNA and protein, and DCFH-DA fluorescence imaging revealed lowered intracellular reactive oxygen species after treatment. The authors are careful to frame these results as associative: the docking evidence is predictive rather than proof of physical binding, the IC50 exceeds concentrations generally achievable in vivo, and icariin is a multi-target compound, so NOX4 should be viewed as a candidate node rather than a confirmed sole mediator.
Limitations temper the excitement but chart a clear path forward. Validation rested on a single cell line, with no animal models or independent clinical tissue testing; ROS measurements were semi-quantitative and would benefit from flow cytometry; and the prognostic model is best understood as a transcriptional surrogate of the NOX4-positive endothelial phenotype. Even so, the study demonstrates the power of stacking network pharmacology on multi-omics and single-cell foundations to move beyond target-list speculation toward mechanistically grounded hypotheses. If genetic perturbation experiments and direct binding assays confirm the icariin–NOX4 axis, the findings could open a route to icariin-based functional food strategies aimed at NOX4/ROS signaling in gastric cancer prevention—and elevate an ancient medicinal herb’s active ingredient into a modern redox-directed research program.
Subject of Research: Multi-omics and single-cell identification of NOX4 as an icariin-associated oxidative stress regulatory node in gastric cancer
Article Title: Integrated Multi‐Omics and Single‐Cell Analyses Implicate NOX4 as an Icariin‐Associated Oxidative Stress Regulatory Node in Gastric Cancer
Article References: Integrated Multi‐Omics and Single‐Cell Analyses Implicate NOX4 as an Icariin‐Associated Oxidative Stress Regulatory Node in Gastric Cancer. (n.d.). https://doi.org/10.1002/fsn3.72390
Image Credits: AI Generated
DOI: 10.1002/fsn3.72390
Keywords: icariin, NOX4, gastric cancer, oxidative stress, network pharmacology, single-cell RNA sequencing, reactive oxygen species, tumor microenvironment, endothelial cells, molecular docking, prognostic model, Epimedium
Cite Scienmag News
Nathaniel Bowman. (October 1, 2026). Icariin Targets NOX4 Oxidative Stress Hub in Gastric Cancer, Multi-Omics Study Finds. Scienmag. https://scienmag.com/icariin-targets-nox4-oxidative-stress-hub-in-gastric-cancer-multi-omics-study-finds/
Nathaniel Bowman. "Icariin Targets NOX4 Oxidative Stress Hub in Gastric Cancer, Multi-Omics Study Finds." Scienmag, 1 October 2026, https://scienmag.com/icariin-targets-nox4-oxidative-stress-hub-in-gastric-cancer-multi-omics-study-finds/. Accessed 1 October 2026.
Nathaniel Bowman. "Icariin Targets NOX4 Oxidative Stress Hub in Gastric Cancer, Multi-Omics Study Finds." Scienmag. October 1, 2026. https://scienmag.com/icariin-targets-nox4-oxidative-stress-hub-in-gastric-cancer-multi-omics-study-finds/

