A novel adrenomedullin receptor signaling-based prognostic model predicts immunosuppressive microenvironment and informs therapeutic stratification in hepatocellular carcinoma

The adrenomedullin receptor signaling pathway plays a crucial role in tumor progression, yet its comprehensive implication in hepatocellular carcinoma (HCC) remains underexplored. This study aimed to develop and validate a multi-gene prognostic signature based on adrenomedullin receptor signaling-related genes (ARGs) and to elucidate its associations with clinicopathological features, immune microenvironment, drug sensitivity, and somatic mutations in HCC. Transcriptomic and clinical data of HCC patients were obtained from The Cancer Genome Atlas and Gene Expression Omnibus databases. Pan-cancer analysis was performed to evaluate the expression and prognostic value of ARGs. A prognostic model was constructed using 116 machine learning algorithm combinations and evaluated by C-index and receiver operating characteristic analysis. The risk score (RS) derived from the model was further correlated with clinical characteristics, immune cell infiltration, drug sensitivity, somatic mutations, and pathway activities. A nomogram was established for clinical applicability. Single-cell RNA sequencing data were analyzed to delineate the cellular heterogeneity of ARG expression within the tumor microenvironment. ADM and RAMP3 were significantly associated with HCC prognosis. The optimal model, built with Random Survival Forest, demonstrated robust predictive performance in both the Cancer Genome Atlas (HR = 9.51, P < .001) and GSE14520 (HR = 2.24, 95% CI: 1.46-3.43, P < .001) cohorts. High-risk patients exhibited advanced T stage, higher histological grade, and poorer overall survival. Computational immune profiling suggested decreased immune cell infiltration and altered immune checkpoint gene expression in the high-risk group. In silico drug sensitivity analysis predicted increased susceptibility to certain targeted agents (e.g., erlotinib) in high-risk patients, though these findings are hypothesis-generating and require experimental validation. Furthermore, RS was weakly correlated with tumor mutational burden (R = 0.16, P = .004), though TP53 mutation frequency did not differ significantly between risk groups after covariate adjustment. The nomogram integrating RS demonstrated favorable predictive accuracy for 1-, 3-, and 5-year survival. Single-cell analysis revealed predominant expression of ARGs in tumor endothelial cells. We developed and validated an ARG-based prognostic signature that effectively stratifies HCC patients into distinct risk subgroups. This model may serve as a hypothesis-generating framework for individualized prognosis prediction and for identifying potential therapeutic vulnerabilities in HCC that warrant prospective investigation.