Abstract
OBJECTIVE: Lymph node metastasis (LNM) critically determines recurrence and survival in cervical cancer (CC), yet current imaging-based methods lack accuracy. Retroperitoneal lymph node dissection leads to many adverse events. This study aimed to develop a clinically actionable molecular signature to predict LNM, enabling personalized surgical planning and improved patient outcomes. METHODS: Transcriptome profiles and clinical data from 193 CC patients, encompassing information on LNM from The Cancer Genome Atlas (TCGA) and an external cohort (GSE26511), were analyzed. The differential expression of mRNAs and lncRNAs was identified using DESeq2. Subsequently, dual machine learning strategies-LASSO regression and the Boruta algorithm-were applied to select robust biomarkers. Finally, the seven-mRNA-lncRNA gene cluster was verified in tumor tissues of CC patients with and without LNM using qRT-PCR. RESULTS: The seven-mRNA-lncRNA gene cluster included four mRNAs (ART3, HRG, MAPT, and SYTL5) and three lncRNAs (AC011239.1, AC125616.1, and RUVBL1.AS1). The expression patterns of the seven DEGs align with their levels in CC tissues. The signature demonstrated high predictive accuracy (AUC: 0.855 in training and 0.807 in testing cohorts). External validation using the GSE26511 dataset confirmed its clinical applicability (AUC: 0.611). Patients with high LNM scores exhibited poorer survival outcomes than those with low LNM scores (p = 0.0034). CONCLUSION: We constructed a reliable prediction model of LNM in CC patients with a seven-mRNA-lncRNA gene cluster. This model guides lymphadenectomy decisions, reduces overtreatment, and enhances patient survival. Our work bridges molecular insights with clinical practice and provides a foundation for further research into the management of CC.