Machine learning-driven multi-omics analysis identifies a prognostic gene signature associated with programmed cell death and metabolism in hepatocellular carcinoma
Abstract Background Hepatocellular carcinoma (HCC) is the most prevalent primary liver malignancy, contributing significantly to global mortality due to limited therapeutic options. Programmed cell death (PCD) and metabolism are key cancer hallmarks, influencing tumor progression and treatment respo...
Saved in:
| Main Authors: | , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
BMC
2025-08-01
|
| Series: | Biological Procedures Online |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12575-025-00286-1 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Background Hepatocellular carcinoma (HCC) is the most prevalent primary liver malignancy, contributing significantly to global mortality due to limited therapeutic options. Programmed cell death (PCD) and metabolism are key cancer hallmarks, influencing tumor progression and treatment response. However, their association in HCC remains insufficiently characterized. Methods We utilized single-cell and bulk transcriptomic datasets to identify differentially expressed genes (DEGs) strongly associated with PCD and metabolism in HCC. Based on prognosis-related DEGs, patients and cells were stratified into high- and low-expression groups using corresponding computational algorithms. The intersecting DEGs from both datasets were analyzed using univariate Cox regression, and a prognostic risk score model was constructed through machine learning algorithms. The model was subsequently evaluated in the context of the immune microenvironment and its relevance to immunotherapeutic responses. Drug repurposing was pursued by integrating machine learning, deep learning, and molecular docking strategies to uncover potential therapeutic options. In parallel, consensus clustering analysis was performed to assess the grouping efficiency of the model-associated genes. Lastly, the expression of the model genes was evaluated in HCC mouse models and cell lines, and the biological function of a representative gene was further investigated through in vitro assays. Results We developed an 18-gene signature based on PCD and metabolism with strong predictive value for overall survival (OS) in HCC patients. Malignant cells with high PCD-Metabolism scores may promote HCC progression by influencing immune infiltration, fibroblast differentiation, and cancer-related pathways. The model also correlated with immunotherapy sensitivity. Leveraging a drug repurposing strategy guided by the PCD-Metabolism model, we identified triazolothiadiazine and fluvastatin as promising compounds targeting RCN2 and CDK4, respectively. Clustering analysis identified two HCC subtypes (C1 and C2), and the subtype enriched with high-risk patients was associated with inferior OS. Notably, CCT3, a key gene in the model, was enriched in tumor regions, and its silencing was found to inhibit the proliferation and migration of HCC cells while regulating ferroptosis- and autophagy-related markers. Conclusion Our study established a PCD-Metabolism-based prognostic model for HCC, offering insights into disease biology and potential avenues for personalized therapy. |
|---|---|
| ISSN: | 1480-9222 |