Multi-omics analysis unveils dysregulation of the tumor immune microenvironment and development of a machine learning-based multi-gene classifier for predicting lateral lymph node metastasis in papillary thyroid carcinoma.

Purpose: Lateral lymph node metastasis (LNM) critically influences surgical decision-making in papillary thyroid carcinoma (PTC). However, the sensitivity of preoperative imageological examination in detecting LNM remains suboptimal, necessitating the development of more accurate diagnostic and predictive tools. This study aims to identify multi-omics biomarkers and construct a predictive model for LNM.

Methods: We performed a comprehensive multi-omics analysis of 50 PTCs presenting with (LNM group) or without lateral lymph node metastases (LNN group) using whole exome sequencing and whole transcriptome sequencing.

Results: Younger age, larger tumor size, and lymphovascular invasion were associated with increased risk of LNM, while invasive follicular subtype was associated with lower risk of LNM. Genomic landscape analysis identified 23 LNM group specific driver mutations and 15 protective variants in the LNN group. Transcriptome analysis identified 444 differentially expressed genes associated with LNM. Weighted gene co-expression network analysis revealed a module that correlated negatively with LNM, with key genes significantly enriched in Notch signaling pathway and Apelin signaling pathway. Notably, elevated neutrophils in tumor immune microenvironment was strongly associated with high LNM risk, suggesting neutrophils as potential early predictors of lateral lymph node metastasis in PTC. A machine learning-based multi-gene classifier was developed to predict LNM, achieving excellent performance with an area under the curve (AUC) of 0.98 in the training set and 0.892 in the test set.

Conclusions: This study provides novel insights into the molecular characteristics of PTC associated with lateral lymph node metastasis, highlighting tumor-infiltrating neutrophils as an independent LNM predictor. The multi-gene classifier developed in this study demonstrates promising clinical utility for improving the accuracy of LNM prediction and guiding personalized treatment strategies in PTC.