Hepatocellular carcinoma (HCC) is a common malignant tumor representing more than 90% of primary liver cancer. This study aimed to identify metabolism-related biomarkers with prognostic value by developing the novel prognostic score (PS) model. Transcriptomic profiles derived from TCGA and EBIArray databases were analyzed to identify differentially expressed genes (DEGs) in HCC tumor samples compared with normal samples. The overlapped genes between DEGs and metabolism-related genes (crucial genes) were screened and functionally analyzed. A novel PS model was constructed to identify optimal signature genes. Cox regression analysis was performed to identify independent clinical factors related to prognosis. Nomogram model was constructed to estimate the predictability of clinical factors. Finally, protein expression of crucial genes was explored in different cancer tissues and cell types from the Human Protein Atlas (HPA). We screened a total of 305 overlapped genes (differentially expressed metabolism-related genes). These genes were mainly involved in “oxidation reduction,” “steroid hormone biosynthesis,” “fatty acid metabolic process,” and “linoleic acid metabolism.” Furthermore, we screened ten optimal DEGs (CYP2C9, CYP3A4, and TKT, among others) by using the PS model. Two clinical factors of pathologic stage (P < .001, HR: 1.512 [1.219-1.875]) and PS status (P <.001, HR: 2.259 [1.522-3.354]) were independent prognostic predictors by cox regression analysis. Nomogram model showed a high predicted probability of overall survival time, and the AUC value was 0.837. The expression status of 7 proteins was frequently altered in normal or differential tumor tissues, such as liver cancer and stomach cancer samples.We have identified several metabolism-related biomarkers for prognosis prediction of HCC based on the PS model. Two clinical factors were independent prognostic predictors of pathologic stage and PS status (high/low risk). The prognosis prediction model described in this study is a useful and stable method for novel biomarker identification.

肝细胞癌（HCC）是一种常见的恶性肿瘤，占原发性肝癌的90％以上。这项研究旨在通过开发新的预后评分（PS）模型来鉴定具有预后价值的代谢相关生物标志物。分析了从TCGA和EBIArray数据库获得的转录组图谱，以鉴定HCC肿瘤样品中与正常样品相比的差异表达基因（DEG）。筛选和功能分析DEGs和代谢相关基因（关键基因）之间的重叠基因。构建了新颖的PS模型以鉴定最佳的特征基因。进行Cox回归分析以鉴定与预后相关的独立临床因素。构建线型图模型以估计临床因素的可预测性。最后，人类蛋白图谱（HPA）在不同的癌症组织和细胞类型中探索了关键基因的蛋白表达。我们筛选了总共305个重叠基因（差异表达的代谢相关基因）。这些基因主要涉及“氧化还原”，“类固醇激素的生物合成”，“脂肪酸代谢过程”和“亚油酸代谢”。此外，我们使用PS模型筛选了十种最佳DEG（CYP2C9，CYP3A4和TKT等）。通过Cox回归分析，病理分期的两个临床因素（P <.001，HR：1.512 [1.219-1.875]）和PS状态（P <.001，HR：2.259 [1.522-3.354]）是独立的预后指标。Nomogram模型显示出较高的总生存时间预测概率，并且AUC值为0.837。7种蛋白在肝癌和胃癌等正常或差异性肿瘤组织中的表达状态经常发生变化。基于PS模型，我们已经鉴定了几种与代谢相关的生物标志物，用于预测肝癌的预后。两个临床因素是病理分期和PS状态（高/低风险）的独立预后指标。本研究中描述的预后预测模型是一种用于新型生物标志物鉴定的有用且稳定的方法。