Identification of an RFC4, CKS2, MCM5 Genes Based Prognostic Signature in Cervical Cancer Using Systems Biology and Machine Learning

Introduction: Cervical cancer makes a significant contribution to the morbidity and mortality of women with scarce prognostic factors. The study was performed to create a good prognostic molecular signature for cervical cancer by utilizing systems-level transcriptome analysis and machine learning methodologies together.
Methods: Gene expression data from TCGA-CESC and GEO were used in the study to find DEGs by GSEA, then WGCNA was done to find disease-related modules. Then, protein-protein interaction analysis was done to discover the most important genes. Machine learning techniques, such as LASSO and Random Forest, were executed to get a minimum prognostic gene set. The prognostic value was determined by Kaplan-Meier survival analysis, time-dependent ROC, and Cox regression. Immune infiltration correlations were ascertained, and 3D protein modeling with molecular docking was utilized for the discovery of potential therapeutic compounds.
Results: A large number of DEGs were able to differentiate cervical carcinoma from normal tissues. The co-expression modules, which were significantly associated with tumors, were enriched in the cell cycle, DNA replication, and immune pathways. The combination of DEG, WGCNA, and PPI analyses revealed RFC4, CKS2, and MCM5 as core genes.
Conclusion: The final model with three genes successfully divided patients into risk high and low groups and also acted as an independent prognostic factor in TCGA-CESC. The expression levels of these genes were related to immune cell infiltration and docking analysis showed good drug-binding affinities, especially for RFC4, thus supporting its characterization as a potential therapeutic target in cervical cancer.