Comprehensive analysis of immune-related gene signature based on ssGSEA algorithms in the prognosis and immune landscape of hepatocellular carcinoma

Background: Hepatocellular carcinoma (HCC) is a malignancy with a poor prognosis. This study aimed to distinguish patients with HCC having distinct tumour immune microenvironment (TIME) features and construct an immune-related gene signature (IRGs) to assess prognosis and provide a basis for personalised therapies. Methods: Transcriptomic data of patients with HCC were downloaded from The Cancer Genome Atlas (TCGA) and Gene Expression Omnibus (GEO) databases. We assessed the immune cell infiltration in each HCC specimen using single sample gene set enrichment analysis (ssGSEA) and classified all patients with HCC into high- and low-immune clusters using a hierarchical clustering algorithm. The ESTIMATE and CIBERSORT computational methods were employed to verify the stability and effectiveness of the immune clusters. Subsequently, the differentially expressed genes (DEGs) of the high- and low-immune clusters and the immune-related genes intersected to obtain the immune-related DEGs. The least absolute shrinkage and selection operator (LASSO) was then employed to screen the optimal genes for the construction of a prognostic predictive signature and to divide patients into high- and low-risk subgroups. The predictive efficacy of the IRGs was further confirmed using Kaplan–Meier survival curves, univariate and multifactorial Cox regression and time-dependent ROC curves in the TCGA and GSE14520 validation cohorts. Furthermore, we developed a nomogram to predict the prognosis. Tumour mutation burden (TMB) was also analysed in the risk groups. Additionally, gene ontology and gene set variation analysis were used for biological function and pathway exploration. Lastly, drug sensitivity analyses were employed to investigate prospective therapeutics in the two risk populations. Results: Immune cluster analysis based on ssGSEA could well distinguish the TIME characteristics of patients with HCC. The stromal score, immune score and ESTIMATE score were all lower in the low-immune cluster. Meanwhile, most of the immune checkpoint-related genes and HLA family genes were overexpressed in the high-immune cluster, suggesting that this cluster could be a beneficial population for immune checkpoint inhibitors therapy. There were 1,617 DEGs between the two immune clusters, of which 414 genes intersected with immune-associated genes. Furthermore, Cox regression analysis revealed 49 DEGs that were associated with survival. Then, 19 DEGs were screened using the LASSO algorithm for IRGs construction and patients were classified into high- and low-risk groups. Both the constructed signature and nomogram had good prognostic predictive efficacy. The signature-based risk score was an independent prognostic predictor in both the TCGA and GSE14520 cohorts. Additionally, there was no significant difference in TMB between the two risk populations. Lastly, the half-maximal inhibitory concentrations of certain chemotherapeutic and targeted therapeutic agents differed between the two risk groups. Conclusion: Our study provides a personalized tool for predicting the prognosis and TIME landscape of HCC and a basis for developing personalised treatment regimens.