An NK cell-related gene signature reveals hepatocellular carcinoma prognosis and immune landscape through integrated single-cell and bulk RNA sequencing

Background The heterogeneity of tumors and complexity of immune microenvironment frequently result in unsatisfactory treatment outcomes for hepatocellular carcinoma (HCC). Natural killer cells (NK cells) are crucial immune cells that exhibit anti-tumor, anti-viral infection, and immune regulatory functions. Nevertheless, dependable biomarkers that utilize NK cell marker genes for personalized treatment and prognostic prediction of HCC remain lacking. Methods A single-cell RNA-seq dataset of HCC was sought from GEO and obtained NK cell marker genes. Differential analysis and Univariate Cox regression were performed on bulk RNA-seq data. Subsequently, HCC samples were divided into different subtypes through consensus clustering. Lasso regression and Multivariable Cox analysis were performed to identify key genes for construction prognostic model. Survival analysis was completed to compare survival differences. The CIBERSORT algorithm was applied to evaluate the immune cell infiltration, and immunotherapy prediction was performed with the assistance of TIDE database. Results A total of 790 NK cell marker genes were identified, and 18 prognosis-related differentially expressed genes (DEGs) were obtained. By consensus clustering, HCC samples were divided into A and B subtypes. The elevated expression of DEGs in subtype A was indicative of a poorer prognosis. Ultimately, prognostic model was constructed using three key genes (AP1S3, RPL23, and HM13), which was validated as an effective predictor of patient risk. Patients in high-risk group had a poorer prognosis. The distribution of certain clinical characteristics varied significantly among different risk groups. Riskscore could be served as an independent prognostic factor, and nomogram consisting of riskscore and clinical characteristics, demonstrated superior predictive capacity. As riskscore increases, the proportion of stromal cells decreases and response to immunotherapy improves. The patients in low-risk group exhibited greater sensitivity to clinically common drugs, suggesting that they are more likely to derive therapeutic benefits from such treatments. Additionally, the expression of genes in the model was confirmed through RT-qPCR experiments and immunohistochemical images. Conclusion The combined analysis of scRNA-seq and bulk RNA-seq has successfully developed a prognostic model that can effectively evaluate the prognostic risk of HCC patients. Preliminary exploration of immune status in different risk groups was performed, which was expected to identify potential targets for HCC immunotherapy.