Analysis and validation of diagnostic biomarkers and immune cell infiltration characteristics in chronic spontaneous urticaria and autophagy based on machine learning

Background: The exact molecular mechanisms governing a heightened risk of chronic spontaneous urticaria (CSU) associated with autophagy remain largely unexplored. Methods: We used the GSE72540 RNA dataset (human biopsy samples) for CSU as our training set, with GSE57178 (human biopsy samples) and GSE72541(human serum samples)as validation sets. We analyzed differentially expressed autophagy-related genes (DEARGs) and explored gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Key genes were identified through protein-protein interactions and machine learning. CIBERSORT helped classify immune cells and ratios. We also created an mRNA-miRNA interaction map and performed Gene set enrichment analysis (GSEA) analysis. Furthermore, real-time quantitative polymerase chain reaction (PCR) (RT-qPCR) was utilized to quantify the expression of potential biomarkers in the peripheral blood of CSU patients and controls. Results: After integrating autophagy genes with differentially expressed genes (DEGs), we identified 32 overlapping genes. whose function enrichment analyses exhibited that they were mainly enriched in NOD-like receptor signaling pathway, tumor necrosis factor (TNF) signaling pathway, autophagy pathway, mitogen-activated protein kinase (MAPK) signaling and immune system. 3 hub genes (MYC, NAMPT, and ERO1L) emerged as potential diagnostic markers via least absolute shrinkage and selection operator (LASSO), SVM-REF, and the random forest method. Immune cell analysis suggests a significant role of neutrophils and mast cells in CSU's progression compared to healthy people. MYC, NAMPT and ERO1L were statistically and positively associated with neutrophils. Additionally, both MYC and NAMPT showed a positive association with mast cells. NAMPT rose consistently in both compartments across GEO and qPCR; MYC surged in skin/our blood yet fell in GEO blood, whereas ERO1L climbed in GEO datasets but not in our blood qPCR. Conclusions: MYC, NAMPT, and ERO1L could potentially be significant contributors to the pathogenesis of CSU and hold promise as novel biomarkers, especially NAMPT, which showed consistently stable expression across different tissues.