Abstract
Mitochondrial-associated programmed cell death signatures (MPCDS) have emerged as critical indicators of cancer progression and treatment response. In hepatocellular carcinoma (HCC), mitochondrial dysfunction and dysregulated PCD pathways are prevalent, yet the application of MPCDS in HCC remains underexplored. This study aimed to identify and characterize the HCC MPCDS, evaluating its prognostic value. We integrated data from TCGA-LIHC and HCCDB18 cohorts, identifying 351 HCC MPCDS genes through differential expression analysis and intersection with known MPCDS genes. We constructed and validated a prognostic model based on 56 out of the 351 genes, regarding their remarkable influence in HCC survival (Cox p-value < 0.0001) using Cox regression and machine learning algorithms. Our analysis revealed distinct clusters with significant differences in survival outcomes, immune landscapes, somatic mutation, and drug sensitivity. The machine learning model demonstrated robust predictive accuracy for HCC prognosis (1-, 3-, 5-year AUC = 0.976, 0.984, 0.976). Clinical HCC samples were collected for experimental validation of the top-ranked genes, KIF2C and CDK4, using qRT-PCR. The results confirmed that the genes were significantly upregulated in tumor lesions. These findings underscore the importance of HCC MPCDS as a prognostic biomarker and highlight its potential for guiding personalized treatment strategies in HCC.