Related Experiment Video
Updated: Jun 23, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
An LSCC-specific R-loop-related model predicts prognosis and neoadjuvant immunotherapy response and identifies
Jie Dai1, Ziyu Qiao1,2, Jia Wang1,2
1Department of Otolaryngology, General Hospital of Northern Theater Command, No.83, Wenhua Road, Shenhe District, Shenyang, 110016, Liaoning, China.
Background:
Laryngeal squamous cell carcinoma (LSCC) lacks robust biomarkers for prognostic stratification and immunotherapy prediction. R-loops have been increasingly implicated in transcriptional dysregulation and tumor progression, but their clinical significance in LSCC remains unclear.
Methods:
We generated an LSCC-specific R-loop-related gene set by intersecting R-loop-associated genes with differentially expressed genes in the TCGA-LSCC cohort, and constructed a prognostic model using LASSO-Cox regression. The model was externally validated in GSE25727 and GSE27020 and further evaluated for neoadjuvant immunotherapy relevance in GSE210287. Single-cell RNA sequencing data from GSE290927 were used to identify the core immune-regulatory gene within the model and explore its potential mechanism in tumor-immune interaction.
Results:
We identified 56 LSCC-specific R-loop-related candidate genes and established a three-gene model that successfully stratified patients into distinct risk groups, with stable performance in external validation cohorts. The model retained prognostic value, distinguished different immune microenvironment states, and was associated with neoadjuvant immunotherapy response. Among the model genes, EIF5A2 showed the greatest consistency with the overall model in terms of high-risk association, poor prognosis, and unfavorable immune-related changes, and was therefore identified as the core gene. Single-cell analysis further showed that EIF5A2 was mainly expressed in tumor cells, while cell-cell communication analysis revealed enhanced interactions between EIF5A2-high tumor cells and exhausted CD8 T cells, with MIF-CD74-related signaling representing the dominant ligand-receptor axis.
Conclusions:
We established an LSCC-specific R-loop-related model with value in prognostic stratification and neoadjuvant immunotherapy prediction, and further identified EIF5A2 as a core immune-regulatory gene within the model. These findings provide a biologically interpretable framework linking R-loop-related tumor states to prognosis, immunotherapy response, and immune remodeling in LSCC.
