Identification of Stress Granule-Related Biomarkers in Childhood Asthma via Integrated Bioinformatics and Machine Learning

Asthma is a common chronic lung disease in children, but the role of stress granules (SGs) in its pathogenesis remains unclear. The present study aims to investigate overlapping genes and regulatory mechanisms between childhood asthma (CA) and SGs via integrated bioinformatics and machine learning. Machine learning algorithms were employed to screen for potential biomarkers among these candidates, followed by expression level validation and ROC analysis. In addition, multiple bioinformatics analyses, such as mRNA-miRNA and TF regulatory network analyses, were also performed to explore the functions of the biomarkers, their regulatory networks. Finally, RT-PCR was employed to validate whether the differences in biomarkers observed between childhood asthma blood samples and healthy control samples consistent with the bioinformatics findings. Through machine learning, 11 candidate genes were initially selected, with 3 biomarkers (HNRNPA2B1, RPE, TAF15) were determined. These biomarkers showed strong diagnostic performance, with AUC all above 0.7. GeneMANIA and Gene Set Enrichment Analysis (GSEA) revealed that their functions were enriched in biological processes such as NADPH regeneration. Immunoinfiltration analysis identified four types of differentially infiltrating immune cells: CD56dim natural killer cell, Central memory CD8 T cell, Immature B cell, and Monocyte. Additionally, RT-PCR validation confirmed significantly elevated mRNA expression of HNRNPA2B1, RPE, and TAF15 in CA patients compared to healthy controls, consistent with the bioinformatics predictions. Our findings highlight HNRNPA2B1, RPE, and TAF15 as promising biomarkers for CA and associated with SG biology, providing new insights into the pathogenesis of CA and identifies potential molecular targets for improved therapeutic intervention.