Design and Validation of a Predictive Model for Hepatocellular Carcinoma Based on Genes With Differential Expression Driven by DNA Methylation.

Background: Hepatocellular carcinoma (HCC) ranks among the world's most lethal cancers, with the majority of cases diagnosed at advanced stages. Accurate prognostic assessment is therefore essential for HCC management. This study utilized DNA methylation (MDGs) and RNA-sequencing data to develop and validate a predictive model for HCC. Methods: MDG profiles, RNA-seq data, and related clinical information were analyzed. Based on the Cancer Genome Atlas (TCGA) dataset, a prognostic signature was developed via univariable and multivariable Cox regression analyses in combination with LASSO regression. Subsequently, a nomogram model was constructed and calibrated using calibration curves. The predictive accuracy of the selected genes was tested through in vitro cellular experiments. In addition, the GDSC dataset was utilized to examine the association between the prognostic signature and drug resistance. Results: Three genes (GLS, TEAD4, and CLGN) were identified and incorporated into the prognostic signature. Low-risk patients exhibited notably improved overall survival (OS) in comparison to high-risk patients. A nomogram model was developed based on clinical variables associated with OS, and its predictive accuracy for OS in individuals with HCC was evaluated via calibration curves. In vitro experiments revealed that the proliferative capacity of cells was notably reduced in the knockout group. The GDSC database was utilized to examine the association between the identified prognostic features and drug resistance. Conclusion: Predictive risk scores were developed based on three candidate MDGs, and a nomogram model was built by integrating clinical variables with these scores. This model can provide personalized prognosis prediction and assess drug resistance among individuals with HCC.