Integrative Single-Cell and Bulk Transcriptomic Analysis Identifies Macrophage-Related Gene Signatures Predictive of...

Background/Objectives: Liver cirrhosis is a major global health challenge and a key risk factor for hepatocellular carcinoma (HCC), a malignancy with high mortality due to late diagnosis. This study aimed to integrate single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing (bulk RNA-seq) data, using single-cell data to identify macrophage-associated transcriptomic changes during the progression from cirrhosis to HCC, and using bulk data to validate these findings in independent cohorts, while developing predictive models for early risk assessment. Methods: We integrated single-cell RNA sequencing (scRNA-seq) and bulk RNA sequencing datasets derived from liver tissues of cirrhosis and HCC patients. Single-cell data were used to identify macrophage subtypes and their dynamic transcriptional changes, while bulk data provided validation in independent cohorts. Gene expression and network analyses were performed, and candidate genes were used to construct diagnostic models with Lasso regression, Random Forest, and Extreme Gradient Boosting (XGBoost). Model performance was evaluated using receiver operating characteristic curves. Results: We identified eleven macrophage-associated genes, among which KLK11, MARCO, CFP, KRT19, GAS1, SOD3, and CYP2C8 were downregulated in HCC, indicating loss of tumor-suppressive and pro-apoptotic functions, while TOP2A, CENPF, MKI67, and NUPR1 were upregulated, reflecting enhanced cell cycle progression, proliferation, and M2 polarization. These are all associated with the progression from liver cirrhosis to HCC. Based on these findings, we established predictive models using Lasso, Random Forest, and XGBoost, which stratified cirrhotic patients into high- and low-risk groups according to cutoff values using liver tissue transcriptomic data. All three models demonstrated high diagnostic performance. Conclusions: This study highlights the critical role of macrophage-associated transcriptomic remodeling in liver disease progression. The machine learning-based predictive models offer a promising approach for early diagnosis and clinical decision-making in patients with cirrhosis.