Machine learning-based screening of asthma biomarkers and related immune infiltration.

INTRODUCTION: Asthma has an annual increasing morbidity rate and imposes a heavy social burden on public healthcare systems. The aim of this study was to use machine learning to identify asthma-specific genes for the prediction and diagnosis of asthma. METHODS: Differentially expressed genes (DEGs) related to asthma were identified by examining public sequencing data from the Gene Expression Omnibus, coupled with the support vector machine recursive feature elimination and least absolute shrinkage and selection operator regression model. Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), Gene set enrichment analysis and correlation analyses between gene and immune cell levels were performed. An ovalbumin-induced asthma mouse model was established, and eukaryotic reference transcriptome high-throughput sequencing was performed to identify genes expressed in mouse lung tissues. RESULTS: Thirteen specific asthma genes were obtained from our dataset analysis (LOC100132287, CEACAM5, PRR4, CPA3, POSTN, LYPD2, TCN1, SCGB3A1, NOS2, CLCA1, TPSAB1, CST1, and C7orf26). The GO analysis demonstrated that DEGs linked to asthma were primarily related to positive regulation of guanylate cyclase activity, gpi anchor binding, peptidase activity and arginine binding. The renin-angiotensin system, arginine biosynthesis and arginine and proline metabolism were the key KEGG pathways of DEGs. Additionally, the genes CEACAM5, PRR4, CPA3, POSTN, CLCA1, and CST1 expression levels were positively associated with plasma cells and resting mast cells. The mouse model revealed elevated nos2 and clca1 expression in the asthmatic mouse group compared with that in normal mice, which was consistent with the findings in asthmatic patients. DISCUSSION: This study identified new marker genes for the prediction and diagnosis of asthma, which can be further validated and applied clinically.