Identification of biomarkers related to lymph node metastasis and prognosis in thyroid cancer based on integrated bioinformatics exploration.

Abstract

Objective: Thyroid Carcinoma (THCA) is a common endocrine malignancy, and Lymph Node Metastasis (LNM) is a critical determinant of poor prognosis. This study aimed to identify LNM-related biomarkers and establish a prognostic model for THCA patients using integrated bioinformatics approaches.

Methods: Clinical data from SEER were analyzed to identify prognostic factors in THCA. Transcriptomic data from TCGA were used to identify differentially expressed genes between LNM and non-LNM samples. Survival analysis, machine learning (Boruta and LASSO), and functional enrichment were applied to screen candidate biomarkers. Immune infiltration and a ceRNA network were also explored.

Results: Six independent prognostic factors were identified from SEER (n = 38,955): age, gender, marital status, race, pathologic T-stage, and pathologic M-stage. A nomogram incorporating these factors showed good predictive performance (AUC = 0.744). Transcriptomic analysis revealed 332 LNM-related genes. Through integrative machine learning, six biomarkers ‒ CLDN10, OTOS, NEXMIF, ACOT6, ASXL3, and SLC4A4 ‒ were selected. These genes were predominantly enriched in immune response and metabolic pathways. M0 macrophages and resting dendritic cells showed significant abundance differences between LNM and NLNM samples, with CLDN10 positively correlated with resting dendritic cells (R = 0.47, adjusted p < 0.01). A ceRNA network involving 14 miRNAs and 4 lncRNAs was constructed, with hsa-miR-15a-5p predicted to regulate SLC4A4, ASXL3, and NEXMIF.

Conclusion: We identified six candidate LNM-related biomarkers and clinical factors associated with THCA prognosis. These findings provide a basis for further investigation into LNM mechanisms and may offer new perspectives for prognostic assessment, though independent validation is required.