Exploring the Diagnosis of Immune-Related Genes in Metabolic Syndrome Based on Three Algorithms

Abstract

Abstract Background The pathogenesis of Metabolic Syndrome (MetS) remains largely unexplored. This study aims to explore the immune-related genes in MetS. Methods The microarray expression dataset GSE98895 was downloaded from the Gene Expression Database (GEO) and the immune-related genes were downloaded from the immune database. The samples of patients with MetS and non-MetS samples were analyzed by CIBORCORT method. The differential expression genes (DEGs) and Immune-related DEGs were extracted. Immune-related DEGs MetS were analyzed by Gene Ontology (GO) and Kyoto Encyclopedia of Genes Genomes (KEGG) pathway enrichment analyses. Protein-Protein Interaction (PPI) network was constructed by string online database and Cytoscape software. We used three algorithms of lasso, SVM-REF, and random forest to screen the attributes of MetS-related differential expression genes and obtained hub genes. These obtained hub genes were utilized to construct the nomogram model. The predictability of each hub gene was also identified by receiver operating characteristic (ROC) curves. The hub genes were then analyzed by GSEA (Gene set Enrichment analysis) and ssGSEA (single-sample Gene Set Enrichment analysis). Results 20 tissue samples from healthy subjects and 20 tissue samples from patients with MetS were obtained. We obtained 946 MetS-related differential expression genes from dataset GSE98895 and 1793 immune-related genes from the immune database. Immune-related genes and MetS-related genes were taken from intersection, and we got 63 immune-related differential expression genes. The expression of dendritic cells and resting mast cells in the samples of MetS had lower expression than those of normal samples. DEGs were mainly enriched in receptor ligand activity, as well as, signaling receptor activator activity by GO analysis. KEGG enrichment analysis indicated immune-related differential expression genes that were enriched in cytokine-cytokine receptor interaction. 13 genes were selected by the LASSO regression analysis (DEFB114, IL19, TNFRSF21, NFYB, CX3CR1, BMP8B, JAG1, DUOX1, IL2, OPRD1, NR1I2, JUN, and MMP9), 10 genes were selected by Random Forest algorithm (IFNG, CX3CR1, TNFRSF21, JUN, MCHR2, MMP9, PGLYRP1, IL1R2, SEMA3F, and CD40 ), and 17 genes were obtained by SVM-REF algorithm(TNFRSF21, JUN, BMP8B, NFYB, DUOX1, DEFB114, NR1I2, IFNG, MMP9, SST, IL2, OPRD1, DEFB103A, GAL, SLIT1, JAG1, SERPIND1). From the intersection of these three algorithms, we obtained three hub genes—JUN, MMP9, and TNFRSF21. The nomogram model of the three hub genes demonstrated good reliability and validity. The predictability of each hub gene was also identified by receiver operating characteristic (ROC) curves, AUC values, all greater than 0.7. GSEA enrichment analysis showed that the up-regulated functions of JUN were mainly concentrated in the amphetamine addition, MMP9 was mainly concentrated in arrhythmogenic right ventricular cardiomyopathy, and TNFSRF21 was mainly concentrated in cocaine addiction. ssGSEA indicated via enrichment analysis that MMP9 was mainly associated with TNFA Signaling via NFKB. In addition, KRAS Signaling, Dn and TNFRSF21 were mainly associated with TGF-β Signaling Pathway and Angiogenesis. Conclusion MMP9, JUN, and TNFRSF21 may be targets for diagnosis and treatment of MetS.