Integrated multi-dataset screening to predict prognosis and identify immunotherapy gene targets in hepatocellular carcinoma patients.

This study systematically predicts prognosis and key gene targets for immunotherapy in hepatocellular carcinoma (HCC) patients based on the joint screening of multiple datasets. Transcriptomic and clinical data from TCGA, GEO, and ICGC were integrated to construct a multi cohort analytical framework. Weighted Gene Co expression Network Analysis was applied to identify key functional modules within the GEO dataset. A comprehensive machine learning ensemble strategy-incorporating over one hundred combinations of classical feature selection and survival prediction algorithms-was employed to derive a robust multi gene prognostic signature. Model performance was evaluated through Kaplan-Meier survival analysis, time dependent ROC curves, and decision curve analysis across multiple independent validation cohorts. Additional analyses examined differential expression across clinical subgroups, immune cell infiltration patterns, immune checkpoint associations, and gene mutation profiles to further elucidate the biological and immunological relevance of the identified genes. Ten key genes were identified. TYMS was identified as a risk factor, while APOL3 and FBXO2 emerged as potential protective factors. Candidate genes were closely associated with features of the immune microenvironment, showing significant correlations with levels of immune cell infiltration and expression of immune checkpoint molecules such as PD-1 and CTLA-4. This study identified core HCC genes with prognostic and immunotherapeutic significance, providing novel targets and a theoretical basis for optimizing risk stratification and personalized treatment.