Sphingolipid metabolism-related genes for the diagnosis of metabolic syndrome by integrated bioinformatics analysis and Mendelian randomization identification

Abstract

Background: The rising global incidence of metabolic syndrome (MetS) highlights the need for more effective diagnostic and therapeutic tools. Sphingolipid metabolites are crucial in MetS pathogenesis, and identifying related biomarkers could improve treatment strategies.

Methods: Differentially expressed genes (DEGs) were extracted from the GSE181646 dataset and compared with sphingolipid metabolism-related genes (SMRGs) to identify differentially expressed SMRGs (DE-SMRGs). Key module genes were obtained via Weighted Gene Co-expression Network Analysis (WGCNA). Machine learning and receiver operating characteristic (ROC) curve validation were used to screen biomarkers, followed by Gene Set Enrichment Analysis (GSEA) and immune cell infiltration analysis. Mendelian randomization (MR) was conducted to explore causal relationships between biomarkers and MetS-related diseases.

Results: A total of 701 DEGs, 599 key module genes, and 30 candidate genes were identified. PTPN18 and TAX1BP3 were validated as biomarkers and were found to be enriched in neuroactive ligand-receptor interactions and vascular smooth muscle contraction pathways. The levels of five immune cell types, including plasmacytoid dendritic cells, exhibited notable differences between the MetS and normal samples. TAX1BP3 exhibited a markedly negative correlation with activated CD8 T cell (r = -0.584), whereas it showed a markedly positive correlation with plasmacytoid dendritic cells (r = 0.744). MR analysis revealed that PTPN18 acted as a protective factor against obesity (P < 0.05, OR = 0.702), hyperlipidemia (P = 0.0015, OR = 0.855), and type 2 diabetes (P = 0.0026, OR = 0.953), but was associated with elevated fasting blood insulin (P < 0.05, OR = 1.036).

Conclusion: PTPN18 and TAX1BP3 were identified as sphingolipid metabolism-related biomarkers for MetS, offering potential promising targets for therapeutic intervention.

Keywords: Key module genes; Mendelian randomization; Metabolic syndrome; PTPN18 and TAX1BP3; Sphingolipid metabolism.

Fig. 1

Identification of differentially expressed genes (DEGs) and the key module genes. A. Volcano map of DEGs. Red dots indicate the top 10 up-regulated genes, while green dots represent the top 10 down-regulated genes. B. Heatmap of top 10 up-regulated and down-regulated gene expressions. C. Venn diagram of differentially expressed sphingolipid metabolism-related genes (DE-SMGRs) via overlap of DEGs and sphingolipid metabolism-related genes (SMGRs). D. Analysis of network topology for various soft-thresholding powers. E. All genes in the GSE181646 were clustered according to the topological overlap matrix (1-TOM). Each branch of the clustering tree represents a gene and co-expression modules were constructed with different colours. F. Correlation heat map of modules and phenotypes. The leftmost color block represents the module, and the rightmost color bar represents the correlation range. G. Scatter plot of turquoise module has the strongest positive correlation with phenotypes in the GSE181646

Fig. 2

Identification and enrichment analysis of candidate genes. A. The Venn diagram of intersecting genes between differentially expressed genes (DEGs) and key module genes. B. Histogram of gene ontology (GO) enrichment analysis of candidate genes (p < 0.05). C. Sankey diagram of Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analysis of candidate genes (p < 0.05)

Fig. 3

Identification and functional enrichment analysis of biomarkers. A. Least absolute shrinkage and selection operator (LASSO) logistic regression algorithm for feature gene selection, B. The support vector machine-recursive feature elimination (SVM-RFE) algorithm was used to analyze the prediction accuracy of the number of different feature genes. The model achieves maximum predictive accuracy when the number of feature genes is 5. C. Venn diagram of candidate biomarkers via intersection between LASSO feature genes and SVM-RFE feature genes. Receiver operating characteristic (ROC) curve of candidate genes (D). Violin maps of the 4 featured genes. E: GSE181646, F: GSE200744, G: GSE145412. H. The KEGG signaling pathways enriched by GSEA analysis for biomarkers (p < 0.05)

Fig. 4

A. The immune cell infiltration status between the MetS group and the normal group was analyzed. The abscissa represented samples from different groups, while the ordinate indicated the immune infiltration scores of different cell types. B. Differences in immune cells between the MetS and normal groups were compared. The abscissa denoted the names of five cell types, and the ordinate showed the immune infiltration scores of each sample in each cell type. Box plots illustrated the maximum, minimum, and median values of the scores. C. A correlation analysis of differentially expressed immune cells was conducted, demonstrating the correlations among the five cell types. The color gradient from red to purple represented the transition from positive to negative correlation. *** indicated P < 0.001. D. A correlation analysis between differentially expressed immune cells and biomarkers was performed. The color gradient from yellow to blue represented a decrease in positive correlation, with *P < 0.05, **P < 0.01, and ***P < 0.001

Fig. 5

The construction of drug prediction and regulatory networks A. The corresponding relationship between hub genes and small-molecule drugs, in which red circles represent hub genes and green rectangles represent small-molecule drugs. B. The TF/miRNA-hub gene regulatory network, in which red circles represent hub genes, purple rectangles represent miRNAs, and blue rectangles represent TFs

Fig. 6

Mendelian randomization (MR) analysis of causal effect of biomarkers on MetS related disease A. Forest plot of MR analysis of PTPN18, TAX1BP3 with MetS related disease. B. Scatter plot of correlation analysis between SNP-exposure factor and SNP-outcome effect. C. Forest plot (Instrumental Variable - Outcome Effect Estimation) to assess the diagnostic performance of exposure factors on outcome at different SNP loci. D. Funnel plot of MR assessment. Each black dot represents a SNP

Fig. 7

The leave-one-out test for Mendelian randomisation (MR). Forest plot of leave-one-out test. The horizontal axis represents the effect of each SNP locus on the outcome through exposure factors, while the vertical axis represents the SNP locus. From top to bottom: fasting blood insulin, hyperlipidemia, obesity, T2DM