ABSTRACT
Hepatocellular carcinoma (HCC) is a major cause of cancer-related death due to early metastasis or recurrence. Tumor angiogenesis plays an essential role in the tumorigenesis of HCC. Accumulated studies have validated the crucial role of lncRNAs in tumor angiogenesis. Here, we established an angiogenesis-related multi-lncRNAs risk model based on the machine learning for HCC prognosis prediction. Firstly, a total of 348 differential expression angiogenesis-related lncRNAs were identified by correlation analysis. Then, 20 of these lncRNAs were selected through univariate cox analysis and used for in-depth study of machine learning. After 1,000 random sampling cycles calculating by random forest algorithm, four lncRNAs were found to be highly associated with HCC prognosis, namely LUCAT1, AC010761.1, AC006504.7 and MIR210HG. Subsequently, the results from both the training and validation sets revealed that the four lncRNAs-based risk model was suitable for predicting HCC recurrence. Moreover, the infiltration of macrophages and CD8 T cells were shown to be closely associated with risk score and promotion of immune escape. The reliability of this model was validated by exploring the biological functions of lncRNA MIR210HG in HCC cells. The results showed that MIR210HG silence inhibited HCC growth and migration through upregulating PFKFB4 and SPAG4. Taken together, this angiogenesis-related risk model could serve as a reliable and promising tool to predict the prognosis of HCC.
ABSTRACT
PURPOSE: The phenomenon of immunogenic cell death (ICD) is intricately linked to numerous antitumor treatments and exerts a profound regulatory function in the tumor immune microenvironment (TIME). We aimed to establish a prognostic signature from the ICD-related biomarkers to differentiate the TIME in hepatocellular carcinoma and predict diverse outcomes for patients with liver cancer. METHODS: ICD score-related genes (ICDSGs) were identified using the weighted gene co-expression network analysis (WGCNA). The ICD score-related signature (ICDSsig) was established by applying LASSO and Cox regression. Model precision was verified using the external datasets. We used independent prognostic variables in clinicopathologic factors to develop a nomogram. Further, clinical characteristics, immune and molecular landscapes, the responses of transcatheter arterial chemoembolization (TACE) and immunotherapy, and chemotherapy sensitivity were analyzed for high- and low-risk patients. RESULTS: ICD score-calculated using the single-sample gene set enrichment analysis (ssGSEA)-displayed strong associations with the TIME in HCC. We identified 34 ICDSGs after integrating the TCGA and GSE104580 datasets. Then, three novel ICDSGs (DNASE1L3, KLRB1, and LILRB1) were screened out to construct the ICDSsig; the prognostic signature performed well in the external databases. The high-risk patients had worse outcomes owing to their advanced pathological state, non-response of TACE, and immune-cold phenotype in the immune landscapes. The immune checkpoint genes, N6-methyladenosine-relevant genes, and microsatellite instability score were increased in the high-risk subgroup, thereby indicating a favorable sensitivity to immunotherapy. Common chemotherapy drugs were more effective in high-risk patients due to low half-maximal inhibitory concentration values. CONCLUSION: The ICDSsig can potentially predict outcomes and therapeutic responses for patients with liver cancer and may assist clinicians in designing individualized treatment strategies.
