Mitotic Hub Gene Network in Colorectal Cancer: Integrated Transcriptomic, Protein-Level, and Clinical-Genomic...

Background: Colorectal cancer (CRC) remains a heterogeneous disease, and improved biomarkers are needed to support prognostic assessment. This study aimed to characterize hub genes in CRC and evaluate whether a gene signature provides biologically meaningful and prognostic information in clinical-genomic models. Methods: We integrated three GEO microarray datasets (GSE110223, GSE110224, and GSE23878) to identify common differentially expressed genes using adjusted p<0.05 and ∣log2FC∣>1. Hub genes and protein expression were identified through protein-protein interaction network analysis using maximal clique centrality and Human Protein Atlas, respectively. Prognostic relevance was evaluated in TCGA-COAD/READ using Kaplan-Meier analysis, multivariable Cox regression, Cox-derived prognostic indices, time-dependent ROC analysis, and regression-based machine learning for internal robustness. Principal component analysis (PCA) was used to derive a standardized PC1-based score from the 10-hub gene signature. Results: A ten-gene mitotic hub signature (TPX2, UBE2C, AURKA, NEK2, PRC1, CCNB1, CDK1, CEP55, FOXM1, and RRM2) was consistently upregulated across the three datasets and enriched for cell-cycle and mitotic pathways. Protein-level and survival analyses supported the biological relevance of several hub genes. In TCGA-COAD/READ, the signature showed limited standalone prognostic value and did not retain independent significance after adjustment for clinical variables, although it contributed modestly in integrated clinical-genomic models. PCA showed a one-dimensional signature, with PC1 capturing the dominant shared expression pattern. Gradient Boosting Regressor (R² = 0.8035, MSE = 0.0473) supported the internal robustness of the DEG-based expression pattern. Conclusions: The ten-gene mitotic hub signature represents a coherent CRC-related proliferative program with limited value as an isolated prognostic marker, but it may still be useful as part of integrated risk models that require external validation.