Abstract
Hepatocellular carcinoma (HCC) is the most prevalent type of liver cancer. Since the tricarboxylic acid cycle is widely involved in tumor metabolic reprogramming and cuproptosis, investigating related genes may help to identify prognostic signature of patients with HCC. Data on patients with HCC were sourced from public datasets, and were divided into train, test, and single-cell cohorts. A variety of machine learning algorithms were used to identify different molecular subtypes and determine the prognostic risk model. Our findings revealed that the risk score (TRscore), based on the genes OGDHL, CFHR4, and SPP1, showed excellent predictive performance in different datasets. Pathways related to cell cycle and immune inflammation were enriched in the high-risk group, whereas metabolism-related pathways were significantly enriched in the low-risk group. The high-risk group was associated with a greater number of mutations of detrimental biological behavior and higher levels of immune infiltration, immune checkpoint expression, and anti-cancer immunotherapy response. Low-risk patients demonstrated greater sensitivity to erlotinib and phenformin. SPP1 was mainly involved in the interaction among tumor-associated macrophages, T cells, and malignant cells via SPP1-CD44 and SPP1-(ITGA5 + ITGB1) ligand-receptor pairs. In summary, our study established a prognostic model, which may contribute to individualized treatment and clinical management of patients with HCC.