Heme metabolism gene-based molecular subtypes and prognostic model predict survival and immune landscape in hepatocellular carcinoma

Metabolism-based molecular subtyping of hepatocellular carcinoma (HCC) has been increasingly investigated. However, the prognostic and immunological relevance of heme metabolism-related transcriptional patterns remains incompletely characterized within an integrated analytical framework. This study aimed to identify heme metabolism-related molecular subtypes, develop a prognostic signature, and examine their associations with immune cell infiltration in HCC. Expression and clinical data from the TCGA-LIHC and GSE76427 cohorts were integrated, comprising 485 patients with HCC. Consensus clustering based on 312 heme metabolism-related genes (HRGs) was used to identify molecular subtypes. Gene set variation analysis was performed to characterize pathway-level differences between the subtypes. A prognostic signature was developed using differential expression analysis, Cox regression, and least absolute shrinkage and selection operator regression. Immune cell infiltration was estimated using CIBERSORT, and its correlations with the risk score and signature genes were assessed. Two HRG-based subtypes were identified: HRGcluster A ( n = 79) and HRGcluster B ( n = 406). Patients in HRGcluster A had significantly poorer overall survival than those in HRGcluster B ( P < 0.001). HRGcluster A was enriched in pathways related to bile acid signaling, fatty acid metabolism, lipid oxidation, and steroid biosynthesis, whereas HRGcluster B was primarily associated with amino acid metabolism. A four-gene prognostic signature comprising G6PD, FTCD, UGT2B15, and CD24 stratified patients into high-risk ( n = 227) and low-risk ( n = 258) groups with significantly different overall survival outcomes ( P < 0.05). The 1-, 3-, and 5-year areas under the receiver operating characteristic curve were 0.650, 0.637, and 0.623, respectively, in the combined cohort; 0.565, 0.626, and 0.592 in the testing cohort; and 0.719, 0.654, and 0.661 in the training cohort. The risk score was negatively correlated with the estimated infiltration of M1 macrophages ( R = − 0.29, P = 0.013) and positively correlated with M2 macrophages ( R = 0.33, P = 0.0073) and neutrophils ( R = 0.27, P = 0.029). Heme metabolism-related transcriptional patterns were associated with differences in overall survival, metabolic pathway activity, and estimated immune cell infiltration in HCC. The four-gene signature provided moderate prognostic discrimination and may complement existing metabolism-based classifications. Further prospective and experimental validation is required before its biological mechanisms or clinical utility can be established.