Abstract
BACKGROUND: Accurate and effective biomarkers for the prognosis of patients with hepatocellular carcinoma (HCC) are poorly identified. A network-based gene signature may serve as a valuable biomarker to improve the accuracy of risk discrimination in patients. METHODS: The expression levels of cancer hallmarks were determined by Cox regression analysis. Various bioinformatic methods, such as GSEA, WGCNA, and LASSO, and statistical approaches were applied to generate an MTORC1 signaling-related gene signature (MSRS). Moreover, a decision tree and nomogram were constructed to aid in the quantification of risk levels for each HCC patient. RESULTS: Active MTORC1 signaling was found to be the most vital predictor of overall survival in HCC patients in the training cohort. MSRS was established and proved to hold the capacity to stratify HCC patients with poor outcomes in two validated datasets. Analysis of the patient MSRS levels and patient survival data suggested that the MSRS can be a valuable risk factor in two validated datasets and the integrated cohort. Finally, we constructed a decision tree which allowed to distinguish subclasses of patients at high risk and a nomogram which could accurately predict the survival of individuals. CONCLUSIONS: The present study may contribute to the improvement of current prognostic systems for patients with HCC.