Identification of dysregulated expressed genes
Three GEO data sets (GSE61145, GSE34198, and GSE66360) were included in this study as the workflow shown in Figure S1. A total of 697, 163, and 734 upregulated and 679, 72, and 741 downregulated genes were obtained with GSE34198, GSE61145, and GSE66360, respectively (Fig. 1A). According to the heat map shown in Figure 1B–D, expression of the top 50 differential genes in the GSE66360 dataset between healthy controls and AMI groups of GSE34198, GSE61145, and GSE66360 showed differential distributions. To identify DEGS, dysregulated expressing genes for GSE34198, GSE61145, and GSE66360 were overlapped, resulting in 134 differentially upregulated and 25 differentially downregulated genes (Figs. 2A, B and Table S2).
Dysregulation expressed genes expressed in the gene expression omnibus (GEO) dataset. (a) Volcanic plots of important dysregulated expressed genes between AMI and healthy control samples. (b–d) Heatmap of the top 50 significantly upregulated or downregulated genes of GSE34198 (b), GSE61145 (c), and GSE66360 (d).
Differentially expressed genes and functional enrichment analysis. (a, b) 134 Differentially upregulated genes (aand 25 differentially downregulated genes (b) Duplicate GSE34198, GSE61145, and GSE66360. (c–e) Top 10 Enrichment Score Values (GO) Values for Heavily Enriched Gene Ontology (GO) Terms, Including Biological Processesc), cellular components (d), and molecular functions (e). (f) Kyoto Encyclopedia of Genes and Genomes (KEGG) Analysis of Differentially Expressed Genes (degs)20, 21, 22.
Function enrichment analysis
Enrichment and kegg pathway enrichment analysis were performed to identify biological functions of degs. The significantly enriched biological functions are summarized in Table S3. Within the biological process category, positive regulation of cytokine production, positive regulation of cell activation, cytokine-mediated signaling pathways, T-cell activation, myeloid leukocyte activation, and acute inflammatory responses were significantly annotated (Fig. 2C). Within the cell component category, secretory granule cavity, cytoplasmic vesicle, vesicle lumen, and tertiary granule cavity were significantly annotated (Fig. 2d). Within the molecular function category, pathways involved in immune receptor activity, cytokine receptor activity, NAD+ Nucleosidase activity and MHC class I receptor activity were significantly annotated (Fig. 2E). KEGG analysis showed that tumor necrosis factor signaling pathway, neutrophil extracellular RAP formation, interleukin (IL)-17 signaling pathway, programmed cell death ligand 1 and programmed death cell protein 1 checkpoint pathway, nuclear factor kappabeta signaling pathway, and Toll-like receptor signaling pathway are important (Figure 2F). These results revealed that DEGs are involved in immunomodulation during the development and progression of AMI.
Identification of hub genes by WGCNA
Based on the representation of the GEO dataset, WGCNA was run to screen important modules and genes that correlate primarily with AMIs. Scale-free indexes and average connectivity were calculated to establish a scale-free network (Fig. 3A,B). Next, a soft threshold of 10 was implemented. The MeyellowGreen module showed the strongest correlation with AMI features (r = -0.36; Figure 3c,d). Scatter plots were constructed for correlation analysis between gene significance of AMI and module membership of Yellow Green Modules, revealing that genes are significantly associated with AMI (Corelation = -0.5, p= 0.0036; Figure 3e).
Hub gene identification by weighted gene coexpression network analysis (WGCNA) analysis. (a) Scale-free index of the soft threshold power (β) of AMI. (b) Average connection analysis of various soft tissue awarding abilities. (c) mRNA clustering dendritic diagram obtained by hierarchical clustering of topology overlap matrix (TOM)-based similarity. (d) Heatmap of correlations between module-specific species between AMI and healthy controls. (e) Correlation analysis between gene significance of AMI and module membership of YellowGreen module.
Identifying hub genes using machine learning
To further reveal the hub gene, 159 degrees were screened by the SVM of GSE34198. The results showed that 39 hub genes were identified with an accuracy of 0.753 (Fig. 4A and Tables S2, S4). These degs were further screened by RF. After determining the MTRY and NTREE parameters, 30 stable genes were retained by RF analysis. This was ranked as an important factor in the splitting of AMIs (Figures 4B, C and Table S2). Lasso analysis confirmed that 14 hub genes were revealed (Fig. 4D, E and Table S2). The hub genes screened by SVM, RF, Lasso, and WGCNA were further overlapped, resulting in 19 hub genes (Fig. 5A). A stepwise regression method was used to further reduce the gene set and finally 10 hub genes, including 10 hub genes. VNN3, fos, IL18RAP, dusp1, rhou, KLHL6, Dusp2, PLA2G7, SLPI and TCN1Identified.
Identification of hub genes using machine learning. (a) Supports Vector Machine (SVM) analysis. (b, c) index (b) and number (c) Cultivated for Random Forest (RF) analysis. (d, e) Cross-validation (LAMBDA) to select the best tuning parameter log (LAMBDA) (d) and the minimum absolute contraction and select operator (Lasso) coefficient profile (e) by Lasso regression analysis.
Development and verification of diagnostic models. (a) cross-regulation of cancelled genes using weighted gene coexpression network analysis (WGCNA), support vector machine (SVM), random forest (RF), and minimal absolute contraction and select operator (Lasso) analysis. (b) Clinical diagnostic model of GSE34198. (c) Clinical diagnostic model of GSE66360.
Development and verification of diagnostic models
dusp1, VNN3 and fos Showed a positive relationship rhoushowed a negative relationship with dusp1, VNN3 and fos(Figure S2). Expression analysis showed that 10 hub genes were dysregulated at GSE34198, GSE66360, and GSE66360 (Figures S3A–C). However, only four hub genes containing VNN3, fos, IL18RAPand dusp1identified in GSE61145. Therefore, GSE34198 and GSE66360 were selected to develop and validate diagnostic models. First, we evaluated the diagnostic values of 10 hub genes for GSE34198 and GSE66360 on the ROC curve. Diagnostic value of VNN3, fos, IL18RAP, dusp1, rhou, KLHL6, Dusp2, PLA2G7, SLPI and TCN1For GSE34198 and GSE66360, 0.767 vs 0.766 vs 0.833, 0.713 vs 0.489, 0.676 vs 0.77, 0.647 vs 0.648, 0.651 vs 0.702, 0.655 vs 0.73, 0.59 vs 0.769, 0.617 Vs. 0.697, respectively (Fig. 5b, c). Nevertheless, we found that the binding model for the 10 hub genes was 0.932 vs 0.953. These results indicate that this combined model may serve as a diagnostic marker predicting AMI and may indicate the involvement of immune cell invasion during development of AMI.
Immune cell infiltration and correlation analysis
To investigate the role of immune cells in AMI, principal component analysis (PCA) was performed according to the expression profiles of 10 hub genes that efficiently distinguish between AMI and healthy control samples (Figures S4A and Table S5). The CiberSort method was applied to analyze the invasion of 22 types of immune cells in clinical samples. Correlation analysis revealed negative regulation between resting NK cells and gamma delta T cells, naive CD4 T cells and M0 macrophages, naive CD4 T cells and neutrophils, CD4 memory T cells and Tregs, and immune cells between neutrophils and CD8 T cells (Figure S4B). The AMI group showed a greater percentage of naive B cells and a percentage of activated CD4 memory T cells and resting mast cells than the healthy control group (p<0.05; Figure S4C).
Expression of VNN3, fos , IL18RAP and dusp1 It was positively correlated with activated CD4 memory T cells, M0 macrophages, and neutrophils, and negatively correlated with CD8 T cells, naive CD4 T cells, Tregs, monocytes, and quiescent mast cells. rhou On the tregg, KLHL6 In plasma cells, monocytes, and resting mast cells, Dusp2 CD8 T cells, PLA2G7 Monocytes and resting mast cells, SLPI With neutrophils and TCN1 Resting CD4 memory T cells, M0 macrophages, and neutrophils were positively correlated. but, rhou Plasma cells, resting NK cells, and M0 macrophages, Dusp2 Naive CD4 T cells and neutrophils TCN1CD4 naive T cells and resting dendritic cells were negatively correlated (p<0.05;Figure 6a). As shown in Figure 6B, HE staining showed that myocardial infarction tissue showed large-scale congestion and edema with tissue necrosis. This indicated that the acute myocardial infarction model was successfully established. Immunohistochemical staining of serial tissue sections was used to detect protein expression of hub genes (fosand IL18RAP), and immune cells. Levels of FOS, IL18RAP, CD4 naive T (CD4), and neutrophils (LY6G) were significantly upregulated with AMI (Fig. 6C). Together, these results suggest that 10 hub genes that regulate immune cell invasion may be potential diagnostic biomarkers of AMI.
Immune cell infiltration of AMI. (a) Heatmap of correlations between 10 hub genes and infiltrated immune cells. Immunohistochemical staining. (b) Control and AMI hematoxylin-eosin staining Vivo. (c) Expression of C-FOS, IL18RAP, CD4, and LY6G by immunohistochemistry. ***Means p<0.001.
