Stemness-Based Gene Signature Predicts Survival and Treatment Response in Liver Cancer

The work integrates single-cell RNA sequencing with bulk transcriptomic data to capture both the cellular makeup of tumors and gene expression measured across whole tissue samples.

Using CytoTRACE analysis, a computational method that estimates a cell’s developmental state, the researchers identified an epithelial subpopulation enriched for stemness characteristics. From the stemness-associated expression patterns, they constructed a seven-gene signature, which they named TSCMS, using LASSO-Cox regression, a statistical technique that selects a compact set of genes while fitting a survival model. When tested across four independent external cohorts, the signature showed favorable prognostic discrimination, meaning it separated patients into risk groups that differed in survival outcomes.

The model also carried biological information. Tumors classified as high risk by TSCMS showed activation of oncogenic and stemness-associated pathways, a higher frequency of TP53 mutations, and an immunosuppressive tumor microenvironment. CellChat analysis, which infers signaling communication between cell types from expression data, revealed extensive predicted bidirectional communication between the stemness-enriched Epi1 epithelial subpopulation and immune cells. The predicted exchange prominently involved MIF signaling originating from the Epi1 cells and VISFATIN/NAMPT signaling from immune populations.

Beyond prognosis, TSCMS was associated with differential response to anti-PD-1 immunotherapy and predicted oxaliplatin sensitivity. The authors emphasize that the signature provides a framework for prognostic stratification and therapeutic characterization rather than a validated clinical decision tool; the findings are associations drawn from transcriptomic cohorts and require prospective testing before any use in patient care.

In laboratory experiments, knocking down ENO1 reduced colony formation, cell migration, sphere formation—a measure of stem-like self-renewal—expression of stemness markers, and resistance to oxaliplatin. Overexpressing ENO1 produced the opposite effects, strengthening the case that the gene contributes to the stemness-linked, drug-resistant behavior the model captures.

In vivo experiments extended the laboratory findings: depleting ENO1 in xenograft models enhanced the antitumor efficacy of oxaliplatin. The researchers also examined an independent clinical cohort of 28 patients with hepatocellular carcinoma receiving combined therapy, where high ENO1 expression in tumors was associated with a poorer treatment response. Because this cohort was small, the clinical association should be interpreted as hypothesis-generating rather than definitive.

Taken together, the study links a computationally derived stemness signature to survival, immune landscape, predicted drug response, and a mechanistically supported resistance gene. Its limitations are those inherent to the design: the prognostic model was validated on existing cohorts rather than tested prospectively, the single-cell and ligand-receptor analyses are predictive rather than directly observed, and the clinical cohort comprised 28 patients. Even so, the work illustrates how combining single-cell resolution with bulk tumor data can yield signatures that connect tumor stemness to both prognosis and therapeutic response in hepatocellular carcinoma.

Subject of Research: A seven-gene tumor stemness prognostic model for survival and therapeutic response prediction in hepatocellular carcinoma

Article Title: A tumor stemness-based prognostic model derived from single-cell and bulk transcriptomics predicts survival and therapeutic response in hepatocellular carcinoma

Article References: Ma, X., Wang, Q., Jiang, J., Hua, Y., Wang, B., Xiao, Y., Xu, W., Zhong, L., Chen, Y., Zhuang, H., Tan, W., & Shang, C. (2026). A tumor stemness-based prognostic model derived from single-cell and bulk transcriptomics predicts survival and therapeutic response in hepatocellular carcinoma. Clinical and Experimental Medicine. https://doi.org/10.1007/s10238-026-02363-5

Image Credits: AI Generated

DOI: 10.1007/s10238-026-02363-5

Keywords: hepatocellular carcinoma, tumor stemness, prognostic model, single-cell RNA sequencing, LASSO-Cox regression, ENO1, oxaliplatin, anti-PD-1 response, TP53 mutation, tumor microenvironment, CellChat, biomarker