Bioinformatic Analysis to Identify a Multi-mRNA Signature for the Prediction of Metastasis in Hepatocellular Carcinoma

Abstract

Hepatocellular carcinoma (HCC) with metastasis indicates worse prognosis for patients. However, the current methods are insufficient to accurately predict HCC metastasis at early stage. Based on the expression profiles of three Gene Expression Omnibus datasets, the differentially expressed genes associated with HCC metastasis were screened by online analytical tool GEO2R and weighted gene co-expression network analysis. Second, a risk score model including 27-mRNA was established by univariate Cox regression analyses, time-dependent ROC curves and least absolute shrinkage and selection operator Cox regression analysis. Then, we validated the model in cohort The Cancer Genome Atlas-liver hepatocellular carcinoma and analyzed the functions and key signaling pathways of the genes associated with the risk score model. According to the risk score model, patients were divided into two subgroups (high risk and low risk groups). The metastasis rate between two subgroups was significantly different in training cohort (p < 0.0001, hazard ratio [HR]: 10.3, confidence interval [95% CI]: 6.827–15.55) and external validation cohort (p = 0.0008, HR: 1.768, 95% CI: 1.267–2.467). Multivariable analysis showed that the risk score model was superior to and independent of other clinical factors (such as tumor stage, tumor size, and other parameters) in predicting early HCC metastasis. Moreover, the risk score model could predict the overall survival of patients with HCC. Finally, most of 27-mRNA were enriched in exosome and membrane bounded organelle, and these were involved in transportation and metabolic biological process. Protein–protein interaction network analysis showed most of these genes might be key genes affecting the progression of HCC. In addition, 3 genes of 27-mRNA were also differentially expressed in peripheral blood mononuclear cell. In conclusion, by using two combined methods and a broader of HCC datasets, our study provided reliable and superior predictive model for HCC metastases, which will facilitate individual medical management for these high metastatic risk HCC patients.