Identification of miRNA biomarkers and development of predictive signatures for early detection and prognosis in cervical precancer and cancer.

backgroundAberrant microRNA expression has been implicated in cervical carcinogenesis, making miRNAs promising biomarkers for early detection and risk stratification of cervical precancer and cancer. This study aimed to identify and validate miRNA expression profiles and signatures that distinguish benign, precancer, and cancer cervical tissues, providing insights into their potential utility in cervical cancer screening and diagnosis.

methodsWe analyzed miRNA expression from archival FFPE cervical tissues of 272 women (100 benign, 89 precancer, 83 invasive cancer) using next-generation sequencing. Differential expression was assessed using DESeq2 with significant miRNAs defined by an absolute log2 fold change ≥ 1 and an adjusted p-value < 0.05. Functional enrichment of miRNA targets was performed using multiMiR and DAVID, and protein-protein interaction (PPI) networks were constructed using STRING and Cytoscape. Machine learning classifiers were developed with recursive feature elimination (RFE) and XGBoost and evaluated in internal and external cohorts. miRNA signatures were further assessed for association with overall survival in the TCGA cervical cancer cohort using multivariable Cox regression models.

resultsNotably, hsa-miR-1246 was the most significant miRNA in cancer vs. benign (log₂FC = 5.58, p = 3.20 × 10⁻41) and cancer vs. precancer (log₂FC = 4.89, p = 5.23 × 10⁻33). hsa-miR-149-5p was most significant in benign vs. precancer (log₂FC = −1.48, p = 1.53 × 10⁻¹⁴). Enrichment analyses highlighted cancer-related pathways with hub genes including TP53, BCL2, STAT3, ESR1, and MYC. RFE with XGBoost identified a 26-miRNA panel for multiclass classification and a 4-miRNA panel for binary cancer vs. benign classification. The binary model achieved strong performance with an AUC of 0.97 and 0.91 in internal and external validation sets respectively. Both the 26-miRNA (p < 0.001) and 4-miRNA (p = 0.01) signatures were associated with poorer overall survival.

conclusionOur findings identify distinct miRNAs and miRNA-based signatures that differentiate benign, precancer, and cancer cervical tissues, highlighting their potential clinical utility as diagnostic and prognostic biomarkers. Furthermore, the development of compact, biologically interpretable miRNA signatures demonstrates strong promise as robust classifiers for early cancer detection and tissue characterization, with possible therapeutic relevance.