What type of study is this?

This is a Observational study (also classified as: Machine Learning).

September 10, 2025Open Access

Multi-omics analysis and metastasis risk factor prediction in N1b stage PTMC: insights into immune infiltration and therapeutic implications

HDHao DaiUniversity of Florida Health QZQian ZhaoJinan University WRWei RenFirst Affiliated Hospital of Xi'an Jiaotong University

Key Points

Two clinical metastasis models were created, with Model A showing superior prediction accuracy (AUC = 0.852).
The neutrophil-to-lymphocyte ratio was identified as an independent risk factor for N1b-stage papillary thyroid microcarcinoma (OR = 2.12, p < 0.01).
Transcriptomic profiling uncovered a robust molecular signature linked to metastatic risk, with an AUC of 0.857 indicating strong predictive capability.
Comprehensive immune profiling indicated diminished CD8+ T cell infiltration, suggesting potential immune evasion in N1b-stage tumors.

Abstract

Background Papillary thyroid microcarcinoma (PTMC) with lateral neck lymph node involvement exhibits a deceptively indolent yet highly invasive phenotype, characterized by early dissemination and slow tumor growth. A comprehensive understanding of integrating multiomics landscapes, circulating immune profiles, and tumor immune microenvironment is essential for more accurate surveillance and tailored therapeutic strategies. Methods Clinical profile and circulating immune-inflammatory markers from 638 PTMC patients were analyzed using multivariate and least absolute shrinkage and selection operator (LASSO) regression to recognize N1b-associated risk indicators. Eight supervised machine learning models were trained via 10-fold cross-validation to select the optimal classifier. Weighted gene coexpression network analysis (WGCNA) alongside machine learning identified metastasis-related gene modules from the integrated RNA-seq profile, leading to a multilayer perceptron gene classifier. Genomic profiling was employed to investigate mutations, copy number alterations, and methylation modifications in signature genes, followed by screening of antineoplastic drugs and docking simulations to explore their therapeutic potential. CIBERSORT, combined with immunohistochemistry, was used to investigate immune infiltration and functional changes in N1b-stage PTMC lesions. Results Two clinical metastasis risk models were developed, with Model A based on the neutrophil-to-lymphocyte ratio (NLR) and Model B on lymphocyte and neutrophil counts, where Model A showed superior generalization (AUC = 0.852) and discriminative performance. NLR was an independent risk determinant for N1b-stage PTMC (OR = 2.12, p 0.01). Transcriptomic profiling revealed a molecular signature ( ALDH1A3 , CTXN1 , MGAT3 , and TMEM163 ) of occult lateral lymph node metastasis, exhibiting strong robustness (AUC = 0.857). Signature genes were predominantly associated with cell adhesion, intercellular signaling, and KRAS dysregulation pathways. Hypomethylation of CTXN1 , MGAT3 , and TMEM163 may underlie transcriptional activation. N1b-stage tumors exhibited reduced CD8+ T and T follicular helper cell infiltration but increased dendritic, γδ T, and activated CD4+ memory T cells, suggesting immune evasion and compensatory immune activation. Discussion This study constructed a robust metastasis prediction nomogram for N1b-stage PTMC and identified metastasis-associated molecular drivers through integrative multiomics analysis. Comprehensive profiling of systemic and tumor-infiltrating immunity revealed key antitumor immune alterations. These findings establish a framework for early metastatic phenotype detection, potentially inspiring relevant immunotherapeutic hypotheses.

AIに質問

Bookmark

View Full Paper