Network pharmacology and cell experiments identify candidate mechanisms for Pei Yuan Yi Fei decoction in non-small cell lung cancer

A study combining network pharmacology, bioinformatics, single-cell RNA sequencing, molecular docking and laboratory experiments identified 95 candidate therapeutic targets associated with Pei Yuan Yi Fei decoction (PYYFD) in non-small cell lung cancer (NSCLC). The results also produced an eight-gene prognostic model with limited predictive capability and cellular evidence that PYYFD altered expression of the modeled genes and reduced viability of PC-9 lung cancer cells.

PYYFD is a 13-herb formulation based on modified versions of the Jin Shui Liu Jun and Huang Qi Gui Zhi Wu Wu decoctions, described in the study as combining strategies of qi tonification and blood activation. Its components include Angelica sinensis, prepared Rehmannia root, Citri reticulatae pericarpium, Pinellia ternata, Poria, raw Astragalus radix, Fritillaria thunbergii bulbus, Cinnamomum ramulus, Paeonia lactiflora radix alba, Gleditsia spina, Hedyotis diffusa and prepared licorice root. The authors noted prior experimental reports of anticancer effects for several individual herbs, but said the mechanism of the complete formula remained insufficiently investigated.

For the computational analysis, the researchers screened TCMSP for compounds with oral bioavailability of at least 30%, drug-likeness of at least 0.18 and half-life of at least 4 hours. This produced 111 active ingredients and 256 corresponding targets. They combined these with 5,912 NSCLC-associated disease targets from GeneCards and OMIM, identifying 203 intersecting genes, and then intersected that set with 6,394 differentially expressed genes from merged lung squamous cell carcinoma and lung adenocarcinoma data. The resulting 95 candidate targets were enriched in Gene Ontology processes involving lipopolysaccharide, bacterial molecules, xenobiotics, oxidative stress, nutrients and chemical stimuli, as well as KEGG pathways including tumor necrosis factor, IL-17, NF-kappa-B, Toll-like receptor signaling, cellular senescence and drug metabolism through cytochrome P450.

The TCGA-NSCLC training cohort included 990 tumor samples after preprocessing. Univariate Cox regression identified 22 survival-associated genes, and LASSO-Cox regression with 10-fold cross-validation retained CDC25C, GSTM2, BIRC5, EPHB2, NUF2, COL1A1, ADH1C and ACHE. Patients divided by the median risk score into high- and low-risk groups showed different overall survival in the training cohort, with a hazard ratio of 1.40 for the high-risk group (95% CI 1.20–1.63; P<0.001). One-, three- and five-year area-under-the-curve values were 0.60, 0.60 and 0.62. The same risk categories showed consistent survival separation in the external GSE31210 and GSE42127 cohorts, whose respective one-, three- and five-year AUC values ranged from 0.61 to 0.70 and 0.64 to 0.69.

Multivariable analysis identified the risk score and age as independent prognostic factors in the training data. Separately from those independent-factor findings, the investigators constructed a nomogram incorporating age and risk score. Its calibration curves were generally consistent with observed one- and three-year outcomes, but its one-, three- and five-year AUC values of 0.609, 0.607 and 0.604 indicated limited predictive capability. The authors also noted that using the same TCGA-NSCLC cohort for feature selection and model training introduced potential optimism bias, reinforcing the need for larger independent validation.

Single-cell analysis of GSE198099 included 20,848 cells from two tumor samples and two paired normal tissues. The cells were assigned to 10 types, including T cells, tumor-associated fibroblasts, monocytes, tumor cells, B cells and natural killer cells. BIRC5 and NUF2 showed relatively high expression in natural killer cells, whereas COL1A1 was preferentially expressed in fibroblasts. COL1A1 expression was significantly higher in tumor-derived fibroblasts than in fibroblasts from adjacent normal tissue. CellChat analysis inferred signaling between fibroblasts and tumor cells, and trajectory analysis separated fibroblasts into EMT-like and myofibroblast states. COL1A1 expression varied across the inferred trajectory and was relatively high in State 2, which harbored the highest differentiation potential; the analysis did not establish that expression peaked there.

Molecular docking selected COL1A1 and quercetin, the active compound with the greatest number of connected targets in the network, for a computational interaction analysis. The predicted binding affinity was −9.0 kcal/mol. This result was presented as a computational prediction rather than direct evidence of binding, and the investigators noted that molecular interaction experiments would still be required.

In laboratory tests, PC-9 cells were exposed to PYYFD at 1, 8 and 16 mg/mL. The CCK-8 assay showed reduced cell viability, with the medium- and high-concentration groups producing the strongest effects at 48 and 72 hours. After 24 hours, viability was 86.91±1.76% at 1 mg/mL and 66.08±3.79% at 8 mg/mL; at 48 hours, it was 51.46±0.47% at 8 mg/mL and 23.88±0.96% at 16 mg/mL. RT-qPCR found decreased mRNA expression of ACHE, NUF2, EPHB2, BIRC5 and CDC25C with increasing concentration, while GSTM2 showed the opposite trend. ADH1C and COL1A1 were reduced at the highest concentration. Western blotting after 48 hours also showed lower COL1A1, NUF2 and BIRC5 protein levels in treated cells than in controls.

Overall, the research provides computational and cell-based evidence about candidate mechanisms associated with PYYFD in NSCLC, not proof of clinical efficacy or a causal pathway. Its main limitations were reliance on public datasets, a small single-cell sample, use of one NSCLC cell line for experimental validation, and the absence of functional experiments establishing that the prognostic genes mediate PYYFD’s effects. The quercetin–COL1A1 interaction remained a docking prediction, and the investigators stated that chemical characterization of PYYFD and further experimental validation were required.

Subject of Research: Computational and cellular investigation of Pei Yuan Yi Fei decoction targets in non-small cell lung cancer

Article Title: Exploring the potential mechanisms of Pei Yuan Yi Fei decoction for non-small cell lung cancer through network pharmacology and experimental validation

Article References: Zhou, Y., Wu, J., Li, M., Jin, X., Liu, S., Yang, J., Shen, H., & Li, F. (2026). Exploring the potential mechanisms of Pei Yuan Yi Fei decoction for non-small cell lung cancer through network pharmacology and experimental validation. Frontiers in Oncology, 16, Article 1938084. https://doi.org/10.3389/fonc.2026.1938084

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1938084

Keywords: non-small cell lung cancer, Pei Yuan Yi Fei decoction, network pharmacology, prognostic modeling, single-cell RNA sequencing, molecular docking, quercetin, COL1A1, cancer cell viability, traditional Chinese medicine, Exploring, potential