Tumor Subtype-Specific Transcriptional Signatures Predict Recurrence Risk in Hepatocellular Carcinoma

Hepatocellular carcinoma exhibits substantial molecular heterogeneity that current clinical staging inadequately captures for predicting post-treatment recurrence. To address this limitation, we developed a survival modeling framework using cell-type-specific transcriptional signatures from RNA sequencing to stratify recurrence risk in HCC patients. Analysis of 1,059 samples across four cohorts demonstrated that tumor immunogenic and proliferative scores satisfied proportional hazards assumptions, enabling construction of a stable multivariable Cox model. The linear predictor showed strong correlation with one-year survival probability, with recurrent patients exhibiting significantly higher risk scores compared to non-recurrent cases. Risk stratification into tertile groups achieved significant survival separation in both training and test datasets. AUC performance analysis revealed optimal discrimination for early recurrence prediction, with maintained but declining accuracy at later timepoints. This biologically-informed risk stratification tool demonstrates robust performance across institutions and may facilitate personalized surveillance strategies for HCC patients following curative treatment.