Integrated Single-Cell and Bulk Transcriptome Analysis Reveals the Prognostic Significance and Immune Regulatory...

Liver Hepatocellular Carcinoma (LIHC) is a high-mortality primary liver cancer. Its treatment and prognosis are highly dependent on disease stage and liver function reserve, necessitating novel biomarkers and optimized therapeutic strategies. Sialylation frequently exhibits abnormal elevation (hypersialylation) in cancers and is recognized as both an important malignant marker and a potential therapeutic target. Transcriptomic, mutational, and clinical LIHC data were procured from TCGA/GEO, extracting sialylation-related genes. Single-cell data underwent quality control, clustering, annotation, and risk-cell subpopulation identification using Seurat/Harmony/SCISSOR. AUCell quantified SRGs activity to identify key differentially expressed SRGs. 10 machine learning algorithms (e.g., SVM, Enet, CoxBoost) were integrated; the optimal StepCox + Enet model was selected via cross-validation, stratifying patients by risk score. The model's clinical utility was validated through GSEA, PPI networks, immune infiltration (CIBERSORT/ssGSEA), and drug sensitivity profiling. This integrated study combined single-cell and bulk transcriptomic data to develop an 11-gene sialylation-related prognostic model for LIHC, demonstrating robust predictive accuracy (AUC > 0.74 across cohorts). High-risk patients exhibited myeloid-driven biology, including enhanced SPP1-mediated cell-cell signaling, TP53 mutations, metabolic dysregulation, and an immunosuppressive microenvironment with elevated TIDE scores. In contrast, the low-risk group displayed active anti-tumor immunity and metabolic homeostasis. Drug sensitivity analysis revealed higher sensitivity to chemotherapeutic agents in high-risk patients. Integrated transcriptomics establishes aberrant sialylation as a key LIHC prognostic biomarker and therapeutic target by stratifying risk subgroups and revealing immunosuppressive microenvironment alterations.