Abstract: Hepatocellular carcinoma is a malignant tumor with high aggressiveness and metastasis. This study aims to construct a predictive model based on metastasis-related genes (MTGs). Cox analysis and Lasso regression analysis were used to build the prognostic model, and the relationship between the risk score of this model and the survival and clinical characteristics of hepatocellular carcinoma (HCC) patients was analyzed. The results of the study constructed a prognostic model containing four MTGs. In the the cancer genome atlas (TCGA) training set, HCC patients in the low-risk group had lower risk scores, fewer deaths and higher 5-year survival (P < 0.01), Cox analysis showed that the model could predict survival of HCC patients independently of other clinical characteristics (P < 0.01), and the AUC values of 1-, 2- and 3-years were all greater than 0.74 in the time-dependent ROC curve prediction analysis. Meanwhile, the validation set also confirmed the above results. Taken together, we constructed a MTGs prognostic model and it was expected to be an independent prognostic factor for prognostic prediction judgment of HCC patients.