Machine Learning Identifies Neutrophil Extracellular Traps-Related Biomarkers for Acute Ischemic Stroke Diagnosis

Haipeng Zhang¹TI WU²Xinghua Li³Shuangqing Liu¹Yuanyuan Wang¹Yang Cao^1*

• ¹Second Hospital of Tianjin Medical University, Tianjin, China
• ²Tianjin Medical University General Hospital, Tianjin, China
• ³Department of Laboratory Medicine, Fengyang College, Shanxi Medical University, Fenyang, China

Purpose: This study aimed to investigate the diagnostic potential of neutrophil extracellular traps (NETs)-related genes in acute ischemic stroke (AIS) through comprehensive bioinformatics analysis. Methods: Two GEO datasets (GSE37587 and GSE16561) were integrated to identify differentially expressed genes (DEGs) between AIS patients and healthy controls. Gene Set Enrichment Analysis (GSEA) was performed to explore functional pathways, while single-sample GSEA (ssGSEA) was used to evaluate immune cell infiltration patterns. NETs-related DEGs (NDEGs) were identified by intersecting the DEGs with previously reported NETS-related genes. Functional enrichment of NDEGs was performed using Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analyses. Key genes were identified via machine learning algorithms, including Least absolute shrinkage and selection operator (LASSO) and random forest (RF). A diagnostic model was constructed based on the identified hub genes and validated using an independent dataset (GSE58294). Potential regulatory miRNAs and candidate therapeutic compounds were predicted using the TargetScan and DSigDB databases, respectively. Results: The discovery dataset included 73 AIS patients and 24 healthy controls, revealed 551 DEGs (225 upregulated, 326 downregulated). The analysis of ssGSEA revealed notable immune dysregulation in AIS patients, characterized by increased neutrophil infiltration and decreased level of Th17, Th1, and TFH cells. GSEA indicated that DEGs were enriched in neutrophil degranulation and innate immune system. NDEGs were significantly enriched in immune regulation and leukocyte apoptosis (GO) and NETs formation pathway (KEGG). Four hub genes—SRC, TLR8, FCAR, and HIF1A—were identified using LASSO and RF algorithms. A diagnostic model based on these genes yielded area under the curve (AUC) values of 0.880 in the training dataset and 0.936 in the validation dataset. Furthermore, three regulatory miRNAs (miR-146a-5p, miR-155-5p, and miR-21-5p) and 23 candidate therapeutic drugs were predicted. Conclusions: To our knowledge, this represents the first comprehensive investigation of NETs-related gene signatures in AIS patients compared with healthy controls. These findings deepen our understanding of immune cell infiltration and the underlying molecular mechanisms involved in stroke, offering novel insights that may enhance diagnostic accuracy and therapeutic strategies for AIS.