Identification of DEGs in AMI versus normal samples
In the present study, we first performed differential expression analysis of 83 AMI samples and 74 normal samples in the metadata dataset. The cut-off criteria were fold change > 3/2 or < 2/3 (|log2(FC)| > 0.5849625) and p value < 0.05. As shown in Fig.1A, a total of 192 DEGs were screened and plotted; green indicates downregulated genes, orange indicates upregulated genes, and gray indicates genes without significant dysregulation (Fig. 1A). The expression levels of DEGs are displayed in the heat map (Fig. 1B).
Construction of the weighted gene coexpression network and identification of modules
We first quantified the distribution of ssGSEA scores for 28 immune cells in the metadata (Fig.2A). To identify immune related genes in the metadata, we selected activated CD4+ T cells, macrophages and regulatory T cells, which are representative of AMI. When the optimal soft-thresholding power for WGCNA was 7, the scale-free topology threshold reached 0.8 (Fig. 2B). The genes were grouped into various modules by hierarchical clustering. Each module represented a group of genes related to expression, and was assigned a specific color (Fig. 2C). The correlation and degree of difference between the gene modules in the three immune cells are shown in Fig. 2D. According to the correlation and p-value in each grid, we finally selected the brown module with the most significant correlation with the three immune cells, and this module contained a total of 387 genes.
Gene ontology (GO) and pathway enrichment analysis
We intersected 192 DEGs with 387 genes in the brown module, and there were a total of 151 overlapping genes (Fig. 3A). The biological functions of these 151 overlapping genes were next assessed via the Matascape database. The 20 most highly enriched terms were related to immune pathways and gene functions (Fig. 3B), and the three most highly enriched terms were myeloid leukocyte activation, the response to lipopolysaccharide and the inflammatory response. Fig. 3C demonstrates the relationships among the top 20 terms in a network plot.
PPI network construction and module analysis
We constructed a PPI network using the STRING database to further explore the interaction relationships among 151 overlapping genes.As shown in Fig. 4, the network contained 135 nodes and 882 edges.
Hub genes verification
The top 5 hub DEGs were determined according to the degree value. The results showed that TLR2 and TLR4 were the most crucial genes, with the highest degree (degree=57), followed by TLR8 at degree=49, IL1B at degree=46, MMP-9 at degree=45, and TYROBP at degree=39. In addition, we used the GSE60993 dataset to verify the expression levels of the 6 hub genes to ensure the accuracy and reliability of the results. Compared with that in controls, the expression of TLR2, TLR4, TLR8, MMP9 and TYROBP in AMI tissues was significantly higher (Fig. 5A-C; Fig. 5E, 5F ; all P <0.05). Although IL1B was also upregulated in AMI, the difference was not significant (P>0.05; Fig. 5D).
Next, we generated ROC curves to establish the diagnostic and prognostic ability of TLR2, TLR4, TLR8, MMP-9, and TYROBP in the metadata and GSE60993 cohorts (Fig. 7). Fig. 7 shows that the area under the curve (AUC) of TLR2, TLR4, TLR8, MMP-9 and TYROBP were 0.763, 0.770, 0.668, 0.834, and 0.646, respectively, in the discovery metadata set. In the validation dataset (GSE60993) , the AUC values of the five genes were 0.849, 0.866, 0.899, 0.882, and 0.807, respectively. Moreover, we merged the five genes into one model and assessed the prediction value in the discovery and validation datasets. The AUC values were 0.883 and 0.958, respectively. These results show that these genes could be considered candidate biomarkers for the AMI diagnosis and prognosis evaluation.
Verification of potential biomarker expression in clinical samples by qRT-PCR
To further determine which genes were of great significance in AMI patients, the expression of the 5 hub genes, TLR2, TLR4, TLR8, MMP-9 and TYROBP, were validated in plasma samples of the experimental group and the control group using qRT-PCR analysis. Consistent with the results of the bioinformatics analysis, the results showed that the expression levels of TLR2 (P-value = 0.0286), TLR4 (P-value = 0.0159), TLR8 (P-value =0.0067), MMP-9 (P-value =0.0467) and TYROBP (P-value = 0.0023) in the plasma of AMI patients were obviously higher than those in the plasma of non-AMI patients (Fig. 6A-E).