Abstract
BACKGROUND: Cancer stem cells (CSCs) play a pivotal role in hepatocellular carcinoma (HCC) pathogenesis, driving tumor initiation, progression, metastasis, and therapeutic resistance. This study aimed to establish a reliable CSC-based signature for HCC through single-cell analysis. METHODS: The study integrates single-cell RNA sequencing (scRNA-seq) data with the Dynamic Data Retrieval Tree (DDRTree) algorithm to identify and characterize CSCs in HCC. A prognostic stemness-associated gene signature was constructed using the Cancer Genome Atlas-Liver hepatocellular carcinoma (TCGA-LIHC) cohort as the training set and validated across two independent HCC datasets (GSE14520-GPL571 and GSE14520-GPL3921). Genes with significant prognostic relevance to CSC biology were prioritized for signature inclusion. The efficacy of model was assessed via survival analysis (Kaplan-Meier) and predictive accuracy evaluation [time-dependent receiver operating characteristic (ROC) curves], demonstrating robust stratification of high- versus low-risk HCC patients and strong prognostic discrimination. RESULTS: A total of 17 CSC-related signatures in HCC were identified by copy number variation (CNV) pattern. Then, we constructed 4-CSCs-related gene predictive model by multivariate Cox regression. The model robustly stratified patients into high- and low-risk cohorts, with high-risk individuals exhibiting markedly reduced overall survival (OS). Kaplan-Meier analysis confirmed significant survival disparity between groups (P<0.001), while time-dependent ROC curves validated the model high predictive accuracy. This prognostic signature may guide the development of personalized therapeutic strategies for HCC. CONCLUSIONS: The prognostic model consisted of 4-CSC-related genes had a prognostic predictive value, providing a new perspective for precision immuno-oncology studies.