**Background:** Hepatocellular carcinoma (HCC) is the sixth most prevalent malignancy and the third leading cause of cancer death worldwide, with approximately 905,677 new cases and 830,180 deaths annually. The 5-year survival rate of HCC is only 5%–30%, and 70% of patients suffer recurrence or metastasis within 5 years after surgery. Pyroptosis is a programmed cell death mediated by the Gasdermin family that releases inflammatory cytokines including IL-1β and IL-18, potentially remodeling the tumor immune microenvironment. While pyroptosis has been implicated in tumor development and immunity, its role in HCC prognosis had not been fully characterized.
**Methods:** Clinical information and mRNA expression data for 1,076 HCC patients were obtained from five public cohorts: TCGA, ICGC, GSE10142, GSE14520, and CPTAC (n=159). Forty pyroptosis-related genes (PRGs) were identified from the Molecular Signatures Database. Unsupervised consensus clustering based on these 40 PRGs was performed in the training cohort (TCGA and ICGC). Differentially expressed genes (DEGs) among pyroptotic clusters were identified (fold change >2, p<0.05). A prognostic risk signature was constructed using LASSO regression with 5-fold cross-validation. The signature was validated in GEO cohorts (GSE10142 and GSE14520) and in the CPTAC cohort at both mRNA and protein levels. Response to sorafenib was assessed in the GSE109211 cohort (n=67). Immune cell infiltration was estimated using CIBERSORT, and drug sensitivity was predicted using oncoPredict based on the CTRP database.
**Key Results:** Gene mutations of PRGs occurred in 54.4% of HCC samples; TP53 was the most commonly mutated gene (30%). Three distinct pyroptotic clusters were identified. Cluster 2 exhibited shorter OS compared to cluster 1 (p=0.003; HR, 1.74; 95% CI 1.26-2.39). A total of 318 overlapping DEGs were identified among the three clusters, from which 24 genes were selected to construct the risk score. In the training set, high-risk patients had significantly shorter OS (P<0.001; HR, 3.06; 95% CI, 2.22-4.24) with a 2-year AUC of 0.77. In the validation set (GSE10142 and GSE14520), high-risk patients also had shorter OS (P=0.008; HR, 1.61; 95% CI, 1.13-2.28) with a 2-year AUC of 0.65. In the CPTAC cohort at the mRNA level, high-risk patients had worse OS (P<0.001; HR, 2.99; 95% CI, 1.67-5.36) with a 2-year AUC of 0.69. At the protein level, a 5-protein model (CYP2A6, CYP2C9, G6PD, FMO3, SEC14L2) with mRNA-protein correlation >0.8 also showed significant OS stratification (P<0.001; HR, 2.97; 95% CI 1.66-5.31) with a 2-year AUC of 0.68. In multivariable analysis, the risk score remained an independent prognostic factor (training: HR=2.82, 95% CI 1.89-4.20, p<0.001; validation: HR=1.70, 95% CI 1.08-2.69, p<0.001). A nomogram integrating risk score, tumor stage, HBV/HCV status, and TMB achieved a 1-year AUC of 0.82. In the sorafenib-treated cohort (GSE109211), the low-risk group had a 45.5% response rate vs. 17.6% in the high-risk group (p=0.02). Predicted IC50 values for forentinib, lenvatinib, linifanib, and sunitinib were significantly lower in the low-risk group.
**Clinical Implications:** This 24-gene pyroptosis-related signature provides a robust tool for risk stratification in HCC patients, independent of traditional clinical variables. The association between low-risk status and better response to sorafenib and other anti-angiogenic drugs suggests potential utility in treatment selection. Validation at both mRNA and protein levels strengthens the clinical translatability of the signature. However, the study is retrospective, and prospective validation with experimental confirmation of the underlying mechanisms is needed before clinical implementation.