Identification of a stemness-related prognostic signature and functional characterization of NCAPG in hepatocellular carcinoma

Cancer stem cells (CSCs) drive drug resistance, recurrence, and heterogeneity in hepatocellular carcinoma (HCC). Although individual stemness-related genes (SRGs) have been explored, comprehensive SRG-based signatures for prognosis and immunotherapy prediction remain underdeveloped. Using an integrated machine learning framework comprising 101 algorithms, we screened SRGs by comparing high- versus low-stemness HCC samples from TCGA-LIHC cohort. The optimal algorithm, StepCox[backward]+Enet[alpha = 0.9], was selected based on the highest average C-index across four independent datasets. A risk stratification system was constructed and validated for survival prediction, immune microenvironment characterization, and immunotherapy response evaluation using TIDE, IPS, TMB analyses, and external ICB therapy cohorts. Functional experiments were performed to validate the role of NCAPG in HCC. A 10-gene SRG signature (CHAF1A, NCAPG, KIF22, KIF4A, PBK, ANLN, RFC4, MCM2, TRAF2, and NSMCE2) was established. Low-risk patients exhibited significantly favorable prognosis across all cohorts ( P < 0.05). The risk score outperformed conventional clinical parameters (average C-index = 0.8015) and 31 published prognostic models. Low-risk tumors displayed an immune-“hot” phenotype with higher immune/stromal scores, elevated CD8 + T cell infiltration, and superior predicted ICB response. High-risk patients showed enrichment of oncogenic pathways and increased sensitivity to multiple chemotherapeutic agents. Knockdown of NCAPG suppressed HCC cell proliferation, migration, invasion, and sphere formation in vitro, and inhibited tumor growth in vivo, accompanied by reduced stemness marker expression. This SRG-based signature provides robust prognostic stratification and immunotherapy response prediction for HCC. NCAPG represents a promising stemness-related therapeutic target.