Background: Cervical cancer is the leading malignancy among women worldwide, posing clinical and public health challenges. This in silico study aims to identify potential diagnostic biomarkers, therapeutic targets, and prognostic markers associated with cervical cancer through integrative bioinformatics approaches. Methods: A hybrid machine learning approach, combining genetic algorithm (GA) and support vector machine (SVM), was applied to high-dimensional gene expression data from publicly available transcriptomic datasets, including the Gene Expression Omnibus (GEO) and The Cancer Genome Atlas (TCGA). A total of 72 Geo samples (Affymetrix, Illumina) served as the primary dataset after normalization. Results: The GA-SVM model achieved about 99% accuracy and AUC with 10-fold cross validation, clearly separating cervical cancer from normal tissues. Eight genes (CXCL9, CTGF, ZNF704, ZEB2, SASH1, PTN, KPNA2, SLC5A1) were identified as diagnostic biomarkers. Protein-protein interaction (PPI) and functional enrichment analyses revealed 42 therapeutic targets (e.g. CDK1, BRCA1, CCNB1, and AURKB) linked to regulating cell cycle, DNA repair, and mitotic processes. Survival analysis identified six genes (CXCL1, DNMT1, MMP1, MYBL2, PCNA, and RRM2) as key prognostic markers. Additionally, transcription factor analysis identified E2F1 and TP63 as major regulators of the prognostic genes, elucidating the molecular mechanisms underlying cervical cancer progression. Conclusion: The identified gene signatures may serve as candidates for hypothesis generation and provide a computational framework to prioritize biomarkers and therapeutic targets in cervical cancer. However, these findings are based on in silico analyses and require experimental and clinical validation before translation into practice.
Farahani et al. (Mon,) studied this question.