Integrative Network Analysis and Machine Learning Identify Cell Adhesion-Related Biomarkers and Immune Heterogeneity in Allergic Rhinitis.
- 1Department of Otorhinolaryngology, Yiwu Traditional Chinese Medicine Hospital, Yiwu, China.
- 2Beiyuan Subdistrict Community Health Service Center, Yiwu, China.
Abstract
Background
Allergic rhinitis (AR) represents a highly prevalent inflammatory disorder of the nasal mucosa, but molecular indicators capable of capturing its disease-specific transcriptional features remain insufficiently defined. This study aimed to identify cell adhesion-related diagnostic biomarkers for AR and to investigate their immune relevance and molecular heterogeneity.Methods
Publicly available AR microarray datasets were analyzed, with GSE75011 serving as the discovery dataset and GSE46171 and GSE40037 used to test the reproducibility of the diagnostic model. A predefined adhesion-related gene set was curated from MSigDB to provide the biological search space for biomarker screening. Genes showing disease-associated expression changes were overlapped with this adhesion gene set to obtain AR-associated adhesion genes. Their biological functions and interaction structure were examined through enrichment analysis, protein interaction mapping, and module extraction. Diagnostic candidates were prioritized by integrating three complementary feature-selection strategies: least absolute shrinkage and selection operator regression, random forest modeling, and support vector machine-recursive feature elimination. The selected markers were then assessed for diagnostic discrimination, individualized prediction, calibration performance, and potential clinical net benefit. In addition, immune infiltration, upstream regulatory networks, candidate compounds, and non-negative matrix factorization-based molecular subtypes were explored.Results
A total of 38 differentially expressed adhesion-related genes were identified in AR, and 13 hub genes were extracted from the key protein-protein interaction module. Integrated machine learning ultimately prioritized TNFSF14 and CD83 as diagnostic biomarkers. The combined TNFSF14/CD83 model showed strong diagnostic separation in GSE75011, with an area under the curve (AUC) of 0.968, and maintained external performance in GSE46171 and GSE40037, with AUC values of 1.000 and 0.758, respectively. Both biomarkers were linked to altered immune-cell features, particularly T-cell-related signatures. Using the two-marker expression pattern, AR samples could be separated into two exploratory molecular patterns that showed differences in immune profiles and enriched biological processes.Conclusions
TNFSF14 and CD83 may represent promising candidate diagnostic biomarkers associated with AR-related immune alterations. These findings also reveal immune heterogeneity within AR and provide a bioinformatics-based framework for biomarker discovery and molecular stratification.
Related
- Integrative Transcriptomics and Machine Learning Identify Temporal Stress Signatures and Astrocytic Neurocan Associated With Hypothalamic Injury.
- Integrative analysis reveals shared biomarkers and pathways in osteoporosis and rheumatoid arthritis using machine learning and single-cell sequencing
- Integrative single-cell and bulk transcriptomic analysis identify R-loop-associated gene signatures driving immune-inflammatory networks in acute pancreatitis
- Integrative genomics and experimental validation identify a mechanosensitive gene signature associated with colorectal cancer liver metastasis
- Integrative transcriptomic analysis and preclinical validation identify lansoprazole as a therapeutic candidate for preeclampsia
BibTeX
@article{he2026integrative,
title = {{Integrative Network Analysis and Machine Learning Identify Cell Adhesion-Related Biomarkers and Immune Heterogeneity in Allergic Rhinitis.}},
author = {Guoshun He and Liying Zhang},
journal = {Immunity, inflammation and disease},
year = {2026},
month = oct,
doi = {10.1002/iid3.70562},
url = {https://europepmc.org/article/MED/42895229},
note = {CC BY},
}
RIS
TY - JOUR
TI - Integrative Network Analysis and Machine Learning Identify Cell Adhesion-Related Biomarkers and Immune Heterogeneity in Allergic Rhinitis.
AU - Guoshun He
AU - Liying Zhang
PY - 2026
DA - 2026/10/01
JO - Immunity, inflammation and disease
DO - 10.1002/iid3.70562
UR - https://europepmc.org/article/MED/42895229
AB - Background Allergic rhinitis (AR) represents a highly prevalent inflammatory disorder of the nasal mucosa, but molecular indicators capable of capturing its disease-specific transcriptional features remain insufficiently defined. This study aimed to identify cell adhesion-related diagnostic biomarkers for AR and to investigate their immune relevance and molecular heterogeneity.Methods Publicly available AR microarray datasets were analyzed, with GSE75011 serving as the discovery dataset and GSE46171 and GSE40037 used to test the reproducibility of the diagnostic model. A predefined adhesion-related gene set was curated from MSigDB to provide the biological search space for biomarker screening. Genes showing disease-associated expression changes were overlapped with this adhesion gene set to obtain AR-associated adhesion genes. Their biological functions and interaction structure were examined through enrichment analysis, protein interaction mapping, and module extraction. Diagnostic candidates were prioritized by integrating three complementary feature-selection strategies: least absolute shrinkage and selection operator regression, random forest modeling, and support vector machine-recursive feature elimination. The selected markers were then assessed for diagnostic discrimination, individualized prediction, calibration performance, and potential clinical net benefit. In addition, immune infiltration, upstream regulatory networks, candidate compounds, and non-negative matrix factorization-based molecular subtypes were explored.Results A total of 38 differentially expressed adhesion-related genes were identified in AR, and 13 hub genes were extracted from the key protein-protein interaction module. Integrated machine learning ultimately prioritized TNFSF14 and CD83 as diagnostic biomarkers. The combined TNFSF14/CD83 model showed strong diagnostic separation in GSE75011, with an area under the curve (AUC) of 0.968, and maintained external performance in GSE46171 and GSE40037, with AUC values of 1.000 and 0.758, respectively. Both biomarkers were linked to altered immune-cell features, particularly T-cell-related signatures. Using the two-marker expression pattern, AR samples could be separated into two exploratory molecular patterns that showed differences in immune profiles and enriched biological processes.Conclusions TNFSF14 and CD83 may represent promising candidate diagnostic biomarkers associated with AR-related immune alterations. These findings also reveal immune heterogeneity within AR and provide a bioinformatics-based framework for biomarker discovery and molecular stratification.
SN - 2050-4527
ER -