Background Hepatocellular carcinoma (HCC) is a top cause of cancer‐related death globally, with late diagnosis due to nonspecific early symptoms. Current single‐factor prognostic models cannot reflect tumor heterogeneity, so a comprehensive tool for risk stratification and personalized treatment is needed. Methods This study employed WGCNA on publicly available datasets (TCGA and GSE54236) to identify core genes associated with chromosomal instability (CIN) in HCC. We initially screened 73 candidate genes, which were then refined to a final set of 20 core genes through an optimization process involving 101 machine learning algorithms. Specifically, the StepCoxboth combined with CoxBoost model was selected as the optimal model, with a concordance index (c‐index) of 0.709. We subsequently developed a multidimensional risk‐scoring model by integrating the expression levels of these core genes with patient clinicopathological parameters and immune cell infiltration data. The model′s performance was evaluated through survival analysis and chemotherapeutic drug sensitivity prediction. Additionally, functional assays were conducted to validate the roles of key genes in promoting the proliferation and invasion of HCC cells. Results The model effectively stratified patients into high‐ and low‐risk groups. High‐risk patients exhibited poorer survival, increased immune cell (particularly T cell) infiltration, higher sensitivity to chemotherapeutics like 5‐fluorouracil and paclitaxel, and a higher TP53 mutation rate. Low‐risk patients were characterized by frequent CTNNB1‐ARID2 comutations and a more active antitumor immune microenvironment. Additionally, SSRP1 and SETDB1 were verified to promote the proliferation and invasion of HCC cells. Conclusion This integrated model, combining genomic and immunological features, is a reliable prognostic tool for HCC patient stratification and personalized chemotherapy, promising for clinical translation and precision medicine in HCC.
Li et al. (Thu,) studied this question.