Identification of Potential Diagnostic Gene Targets for Pediatric Sepsis Based on Bioinformatics and Machine Learning

Abstract

Purpose: To develop a comprehensive differential expression gene profile as well as a prediction model based on the expression analysis of pediatric sepsis specimens. Methods: In this study, compared with control specimens, a total of 708 differentially expressed genes in pediatric sepsis (case–control at a ratio of 1:3) were identified, including 507 up-regulated and 201 down-regulated ones. The Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of differentially expressed genes indicated the close interaction between neutrophil activation, neutrophil degranulation, hematopoietic cell lineage, Staphylococcus aureus infection, and periodontitis. Meanwhile, the results also suggested a significant difference for 16 kinds of immune cell compositions between two sample sets. The two potential selected biomarkers (MMP and MPO) had been validated in septic children patients by the ELISA method. Conclusion: This study identified two potential hub gene biomarkers and established a differentially expressed genes-based prediction model for pediatric sepsis, which provided a valuable reference for future clinical research.