Integrated bioinformatics analysis reveals cross-talking hub genes and therapeutic agents between sepsis and acute myocardial infarction

Abstract

Background

Sepsis and acute myocardial infarction (AMI) are two significant diseases that may share overlapping etiological mechanisms. This study aims to systematically identify core genes common to both conditions and to explore their potential as therapeutic targets and drug candidates through an integrative analysis of clinical data and bioinformatics.

Methods

The AMI dataset was obtained from the GEO database, and RNA sequencing data were collected from blood samples of patients with sepsis at our hospital. Common genes were identified using differential expression gene analysis (DEG) and weighted gene co-expression network analysis (WGCNA). Functional enrichment analyses, including Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, were performed. A protein-protein interaction (PPI) network was constructed, and hub genes were identified using the MCC/Degree algorithm. Diagnostic value was assessed via receiver operating characteristic curve analysis. Immune infiltration patterns, single-cell sequencing data, and molecular docking simulations were employed to evaluate immune relevance and identify potential therapeutic compounds.

Results

A total of 417 genes were identified between sepsis and AMI, with enrichment analysis revealing significant involvement in inflammatory responses. Three hub genes—JAK2, MYD88, and TIMP1—were selected for further investigation. ROC curves confirmed their strong diagnostic performance for both diseases. Immune infiltration analysis showed that these core genes were significantly correlated with the infiltration levels of various immune cell types. Molecular docking indicated that quercetin exhibited stable binding affinity with the proteins encoded by these genes. qPCR validation further confirmed the upregulation of these three genes, supporting the anti-inflammatory effects of quercetin as a potential targeted therapy.

Conclusion

JAK2, MYD88, and TIMP1 were identified as shared core genes in sepsis and AMI. These genes not only serve as potential diagnostic biomarkers but also offer novel targets for developing common therapeutic strategies for both conditions. Furthermore, quercetin emerges as a promising candidate for targeted treatment.

1. Introduction

Sepsis, a systemic inflammatory response syndrome triggered by infection, remains a primary contributor to global morbidity and mortality in intensive care units (ICUs) [1]. With approximately 50 million cases and 11 million deaths annually, it represents a profound threat to public health [2]. The pathogenesis of sepsis is driven by profound immune dysregulation. While an initial hyper-inflammatory surge aims to eliminate pathogens, the subsequent loss of immune homeostasis often results in extensive tissue damage. Conversely, a compensatory anti-inflammatory response can precipitate immunosuppression, heightening the risk of secondary infections and multi-organ failure. Consequently, early diagnostic precision and timely therapeutic intervention are essential to improve clinical outcomes and mitigate the associated socioeconomic burden [3]. A central regulatory hub in this process is the NF-κB signaling pathway; its activation drives the pathological release of pro-inflammatory cytokines, including TNF-α, IL-6, and IL-1β [4]. Elucidating the molecular mechanisms governing these pathways is vital for developing targeted therapeutic strategies.

Similarly, AMI involves a complex interplay of ischemic injury and immune-mediated remodeling. Triggered by atherosclerotic plaque rupture and subsequent coronary occlusion, AMI initiates a biological cascade orchestrated by both cardiac and immune cell populations [5]. Early ischemia promotes the release of damage-associated molecular patterns (DAMPs), which bind to pattern-recognition receptors (PRRs), activating the complement system and inducing reactive oxygen species (ROS) production [6]. This pro-inflammatory environment facilitates the infiltration of leukocytes into the infarcted myocardium, further exacerbating tissue damage. As the condition transitions into the reparative phase, macrophages shift toward an anti-inflammatory phenotype to promote tissue restructuring and inflammation resolution [7,8]. Given that macrophages persist within the myocardium for months and fundamentally influence both initial injury and long-term remodeling, modulating these immune responses is a critical frontier for AMI intervention.

Recent advancements in bioinformatics have enabled the exploration of the systemic links between sepsis and AMI. This study aims to identify shared diagnostic biomarkers and therapeutic targets through integrated computational analysis. Utilizing gene expression profiling and Protein-Protein Interaction (PPI) networks, we will characterize the shared molecular landscape of both conditions. We will employ the ImmuCellAI to quantify immune cell infiltration and the DGIdb platform to identify potential drug candidates. Furthermore, single-cell RNA sequencing (scRNA-seq) analysis will be conducted to resolve the expression profiles of key genes across specific immune cell subsets in sepsis populations. Finally, structure-based virtual screening and molecular docking will be utilized to predict the binding affinities of small molecules to identified core proteins. By integrating these multi-omic and computational approaches, this research seeks to elucidate the common etiology of sepsis and AMI, providing a theoretical foundation for novel, cross-disease therapeutic strategies.

2. Methods

2.1. Study design flowchart

Figure S1 presents a detailed schematic of the investigative methodology employed in this study. To identify shared hub genes between AMI and sepsis and screen potential targeted drugs, an integrated multi-omics approach was adopted. This comprehensive strategy combined meta-analysis with single-cell RNA sequencing techniques, alongside computational assessments through molecular docking. The application of these diverse methods enabled a systematic investigation of the primary targets and fundamental mechanistic pathways of the common hub genes shared by both diseases.

2.2. Clinical data and transcriptome data preprocessing

Clinical data for this study were derived from blood samples collected from 23 sepsis patients and 10 healthy individuals between January 2019 and December 2019 in the Emergency ICU of Southwest Medical University Hospital. Sample size sufficiency was determined using PASS 15.0 software, ensuring a power of 0.80 with an alpha of 0.05 based on preliminary pilot data. Peripheral blood samples were collected in EDTA tubes. Total mRNA was extracted using the RNeasy Mini Kit (Qiagen, Germany). RNA integrity was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies), requiring an RNA Integrity Number (RIN) ≥ 7.0 for all samples. Sequencing libraries were constructed using the Illumina TruSeq Stranded mRNA Library Prep Kit. Libraries were sequenced on the Illumina HiSeq X Ten platform with a 150 bp paired-end strategy, achieving a sequencing depth of at least 30 million reads per sample. Raw data underwent quality control using FastQC and were trimmed via Trimmomatic to remove adapters and low-quality bases (Q < 20). Clean reads were mapped to the human reference genome (GRCh38) using STAR. Gene expression levels were quantified as fragments per kilobase per million mapped reads (FPKM). The dataset generated is accessible in the CNGBdb (Accession: https://db.cngb.org/CNP0002611). The enrollment criteria for study participants included 1. a diagnosis of sepsis and admission to the EICU; 2. fulfillment of the diagnostic criteria for sepsis version 3 as published by the Society of Critical Care Medicine (SCCM) and the European Society of Intensive Care Medicine (ESICM) in 2016; 3. age between 16 and 65 years; 4. provision of written informed consent by clinical participants; and 5. adherence to the principles outlined in the Declaration of Helsinki. The study protocol received approval from the Ethics Committee of the Affiliated Hospital of Southwest Medical University (Ethics Approval No. ky2018029) and was registered with the China Clinical Trial Registry (Registration No. ChiCTR1900021261). Exclusion criteria comprised (1) patients with pre-existing organ failure, (2) a prior history of immune system disorders, (3) a prior history of hematologic disorders, and (4) individuals who declined to participate in the study. The study protocol adhered to all principles outlined in the Declaration of Helsinki.

Following the established literature selection protocol, the transcriptomic dataset GSE249812 [9] was selected based on the following criteria: (1) samples derived from human PBMCs; (2) inclusion of acute phase samples and appropriate recovery controls; (3) availability of complete clinical metadata. To ensure data comparability, the expression matrix was first transformed using log₂(x + 1) to stabilize variance and handle zero values. Subsequently, quantile normalization via the ‘normalizeBetweenArrays’ function in the ‘limma’ package was performed. This dual-normalization strategy was employed because, while log₂ transformation addresses the skewness of gene expression distribution, quantile normalization is essential to eliminate inter-sample technical variability and batch effects, thereby fulfilling the distributional assumptions required for the subsequent empirical Bayes linear modeling.

2.3. WGCNA network construction and module identification

WGCNA is a systems biology approach employed to characterize gene correlation patterns across various samples. This method enables the identification of gene sets that display a high degree of co-expression. By analyzing the connectivity within these gene sets and their associations with phenotypes, WGCNA can pinpoint candidate biomarker genes or therapeutic targets [10]. The OEBIOTECH platform (https://cloud.oebiotech.com/task/) and the WGCNA R package were utilized for network construction. The input data consisted of the top 5,000 genes ranked by median absolute deviation (MAD) to ensure computational efficiency and signal stability. A similarity matrix was constructed based on the Pearson correlation coefficients between gene expression profiles. To achieve a scale-free topology (scale-free fitting index R² > 0.85), a soft-thresholding power (β) was selected. The adjacency matrix was transformed into a Topological Overlap Matrix (TOM) to minimize noise and spurious correlations. We constructed a signed weighted network to preserve the direction of gene correlations. To exclude false positives, a consensus analysis with 1,000 permutations of random data was performed to validate the robustness of the identified modules. Module construction involved hierarchical clustering and dynamic tree-cutting (minimum module size = 30). Module membership (MM) and gene significance (GS) were assessed; the module with the highest Pearson correlation and an adjusted p-value < 0.05 was designated the hub module.

2.4. GO and KEGG enrichment analysis

GO and KEGG pathway enrichment analyses were conducted on the DEGs (criteria: |log2FC| > 1, FDR < 0.05) using the Microbiotics platform. Enrichment analyses employed the hypergeometric distribution test, with the Benjamini-Hochberg method used for multiple testing correction. Significant enrichment was defined as a corrected p-value (FDR) < 0.05.

2.5. PPI network construction

417 potential shared driver genes were imported into the STRING database (https://cn.string-db.org/). An interaction score threshold of > 0.400 was applied to construct the network. PPI networks were visualized in Cytoscape. The top 10 genes were identified using the MCC algorithm to pinpoint core components.

2.6. Immune infiltration analysis

Correlations between clinical data and the abundance of 24 immune cell types were analyzed by uploading a list of normalized gene expressions to the ImmuCellAI platform (http://bioinfo.life.hust.edu.cn/web/ImmuCellAI). ImmuCellAI (Immune Cell Abundance Identifier) is a tool designed to infer the relative proportions of 24 immune cell types from transcriptomic data generated by microarray or RNA sequencing, specifically for normal tissue or blood samples. The abundance of immune cells was predicted by calculating ssGSEA enrichment scores based on the expression bias profiles of each cell type. Subsequently, correlations between core biomarkers and the expression levels of infiltrating immune cells were assessed using Spearman correlation analysis.

2.7. Meta-analysis

To comprehensively analyze the transcriptional differences in the expression of crossover genes across diverse populations, this study acquired data from the Gene Expression Omnibus (GEO) public repository. Ten datasets of sepsis peripheral blood samples were retrieved with the following accession numbers: GSE163151, GSE232753, GSE95233, GSE69063, GSE243217, GSE236713, GSE185263, GSE154918, GSE134347, and GSE100159. Subsequently, the collected datasets were standardized through log2 transformation to ensure data consistency. The standardized data were then segregated into sepsis (Sepsis) and normal (NC) groups. Individual effect sizes (Standardized Mean Difference) were calculated for each gene. A random-effects model was employed if heterogeneity (I²) exceeded 50%; otherwise, a fixed-effects model was used. MultiMeta analysis was performed to assess gene expression consistency.

2.8. Single-cell sequencing

Due to the limited availability of PBMC sequencing data for AMI samples, a sepsis-specific single-cell dataset was employed for analysis in this investigation. A total of 12 peripheral blood samples—comprising 7 from healthy individuals and 5 from sepsis patients—were processed to prepare single-cell suspensions. Libraries were generated using the 10× Genomics Chromium Next GEM Single Cell 3′ Reagent Kit (v3.1) and subsequently sequenced on the Illumina NovaSeq 6000 platform. Raw FASTQ files were processed using Cell Ranger (v7.0) for alignment to the reference genome and UMI counting. For downstream analysis, the Seurat R package was utilized for enhanced quality control and data processing. Data filtering was strictly performed to exclude low-quality cells based on the following criteria: cells with fewer than 200 or more than 2,500 detected genes, or those with a mitochondrial gene content exceeding 10%. Following log-normalization and the identification of highly variable genes, dimensionality reduction was conducted using Principal Component Analysis (PCA), followed by t-SNE or UMAP visualization. Cell type annotation was initially performed using the ‘SingleR’ package integrated with the Human Primary Cell Atlas reference and further refined through manual curation of canonical markers. To identify characteristic disease-associated cells, the ‘FindAllMarkers’ function was employed to determine marker genes across different clusters, which were visualized using VlnPlot and FeaturePlot to illustrate expression disparities from multiple perspectives.

2.9. Screening and molecular docking of drug candidates

Utilizing the DSigDB library [11,12] within the Enrichr platform (https://maayanlab.cloud/Enrichr/), we screened drug candidates with significant corrected P-values by calculating P-values and binding scores for core hub genes. These prospective small molecule compounds are anticipated to function as combination therapeutic agents for AMI and sepsis. The 2D structures of the small molecule ligands were retrieved from the PubChem database (http://pubchem.ncbi.nlm.nih.gov/) and imported into ChemOffice 20 software to generate 3D structures, which were subsequently saved as mol2 files. We identified the crystal structures of protein targets with higher resolution as molecular pair acceptors using the RCSB PDB database (http://www.rcsb.org/). The proteins were pre-processed through de-watering and de-phosphorylation using PyMOL 2.6 software and saved as PDB files. Molecular Operating Environment (MOE) 2019 software was utilized to minimize the energy of the compounds and to preprocess the target proteins for the identification of active pockets. Subsequently, molecular docking was conducted using MOE 2019, with the number of docking operations set to 50. Binding activity was evaluated based on the magnitude of binding energy and visualized using PyMOL 2.6 and Discovery Studio 2019 software.

2.10. Main reagents and cell lines

The mouse monocyte macrophage cell line RAW264.7 and the rat cardiomyoblast cell line H9c2 were obtained from Shenzhen Youli Biotechnology Co. Quercetin was acquired from Chuangrong Science and Technology Co. The CCK-8 kit (No. K1018) and TRIzol reagent were sourced from Luzhou Xinke Sheng Biotechnology Co.

2.11. Cell culture and model construction

RAW264.7 and H9c2 cells were cultured in DMEM high-glucose medium supplemented with 10% fetal bovine serum and 1% penicillin-streptomycin at 37 °C with 5% CO₂. To induce the hypoxia/reoxygenation (H/R) model, H9c2 cells were seeded in culture plates overnight until reaching 70–80% confluency. Subsequently, the cells were switched to sugar-free, serum-free DMEM and exposed to hypoxia (94% N₂, 5% CO₂, 1% O₂) for 6 h to mimic ischemia. Following hypoxia treatment, the cells were returned to normal complete medium and placed in a normoxic incubator (5% CO₂, 21% O₂) for 24 h for reoxygenation.

2.12. Cell viability detection

Cell viability was assessed using the Cell Counting Kit-8 (CCK-8) according to the manufacturer’s instructions. Following a 24-hour reoxygenation period, the medium was carefully removed from the culture plate, and 100 μL of fresh complete medium containing 10% CCK-8 reagent was added to each well of a 96-well plate. The plates were incubated at 37 °C for 2 h in the absence of light. Absorbance was measured at 450 nm using an enzyme meter. The cell viability of the control group was designated as 100%, and the viability of the H/R group was expressed as a percentage relative to the control group.

2.13. RNA extraction, reverse transcription and quantitative real-time PCR(qPCR)

Total RNA was isolated from cell groups using the TRIzol method. Subsequently, 1 μg of total RNA underwent reverse transcription to produce cDNA, following assessment of RNA concentration and purity with an enzyme marker. The resulting cDNA served as a template for qPCR, employing a FastReal QPCR Premix Reagent (SYBR Green) cartridge from TIANGEN. The qPCR protocol consisted of an initial denaturation step at 95 °C for 2 min, followed by 40 cycles at 95 °C for 5 s and 60 °C for 15 s. A final extension was performed at 95 °C for 2 min. The relative expression level of the target gene mRNA was determined using the 2^(–ΔΔCt) method, with GAPDH serving as the internal reference gene. The qPCR primers utilized in this study were supplied by Beijing Prime Biotechnology Co. Details regarding the primer sequences are presented in Table 1.

Table 1.

Primer sequences for qPCR.

Gene names	Primer sequences (5′ – 3′)
GAPDH	Upstream: CAATGAATACGGCTACAGCAAC downstream: AGGGAGATGCTCAGTGTT
TNF-α	Upstream: CTTGTTGCCTCCTCTTTTGCTTA downstream: CTTTATTTCTCTCAATGACCCGTAG
IL-6	Upstream: TCACAGAAGGAGTGGCTAAGGACC Downstream: ACGCACTAGGTTTGCCGAGTAGAT
MYD88	Upstream: TGGCCCTGGTATGTAGTCTC downstream: CCTCAGTCTCAGGTAGGTAGA
JAK2	Upstream: TGGAGTATGTTTCTGTGGAGAC downstream: TAATTTAAAACCAAATGCTTGTG
TIMP1	Upstream: GAGACACACCAGAGCAGATACC downstream: TGGTCTCGTTGATTTCTGGGG

2.14. Statistical analysis

All experiments were conducted independently a minimum of three times. Data are presented as mean ± SD. The normality of the data distribution was assessed using the Shapiro-Wilk test. For normally distributed data, parametric tests were utilized due to their higher statistical power in detecting true effects within biological replicates. One-way ANOVA with Tukey’s post-hoc test was employed for multiple groups, while Student’s t-test was used for two groups. A p-value < 0.05 was deemed statistically significant (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001).

3. Results

3.1. Identifying differentially expressed genes in AMI and sepsis

To delineate the transcriptional alterations in both conditions, DEGs were identified using the ‘limma’ R package, applying a rigorous filtering kinship of |log₂(Fold Change)| > 1 and an adjusted p-value < 0.05. Using the AMI dataset (GSE249812), we identified 287 DEGs, comprising 203 up-regulated and 84 down-regulated genes (Figure S2A). Volcano plots were employed to visualize the global distribution of these DEGs, while heat maps illustrated the expression patterns of the most significantly up- and down-regulated genes (Figure S2B). Similarly, applying the same statistical criteria to the integrated sepsis transcriptomic data, we identified 16,048 DEGs, including 7,448 up-regulated and 8,600 down-regulated genes (Figure S2C, D). To explore the shared molecular basis between the two diseases, a cross-analysis was performed. We identified a total of 53 overlapping DEGs common to both the AMI and sepsis datasets (Figure S2E). The significance of this gene overlap was evaluated using a hypergeometric distribution test, which yielded a significant p-value (p < 0.001), indicating that the intersection of these genomic signatures is highly unlikely to have occurred by chance

3.2. Weighted co-expression network analysis (WGCNA) for AMI and sepsis

To further elucidate the functional gene clusters associated with the progression of both AMI and sepsis, we performed WGCNA on the AMI dataset (GSE249812) and our self-collected sepsis cohort. Sample clustering analysis confirmed the absence of significant outliers, ensuring the high quality of the input expression matrices. In the process of constructing the adjacency matrix, the soft-thresholding power (β) was determined to be 30 for both datasets to approximate a scale-free topology (Figures S3A, B). Notably, although the traditional scale-free topology fitting index (R²) did not fully exceed the conventional 0.8 threshold, this phenomenon is frequently observed in clinical transcriptomic data characterized by high biological heterogeneity and complex noise profiles. After a systematic evaluation of network connectivity and topology, a power of 30 was meticulously selected based on the following rigorous criteria: (1) Optimization of Network Sparsity: A higher power was employed to effectively suppress weak, non-specific background correlations, thereby ensuring that only the most robust co-expression relationships were retained. (2) Biological Signal Amplification: By increasing the stringency of the soft-thresholding, we focused the analysis on high-confidence ‘hub’ signals, which is essential for identifying conserved regulatory mechanisms across two distinct disease states. (3) Methodological Convergence: Utilizing an identical, stringent power for both AMI and sepsis datasets facilitated a more reliable downstream intersection analysis by maintaining consistent network density. Under these parameters, we identified 6 highly stable modules in the AMI dataset and 11 modules in the sepsis dataset (Figure S3C, D). Despite the Despite the relatively streamlined number of modules, they represented the most biologically coherent expression patterns. Correlation analysis between module eigengenes and clinical phenotypes revealed that the plume2 module exhibited the strongest positive correlation with AMI (r = 0.68, p < 0.001; Figure S3E), while the salmon module demonstrated the most significant positive association with sepsis (r = 0.81, p < 0.001; Figure S3F). Furthermore, a robust correlation between Gene Significance (GS) and Module Membership (MM) was observed (r = 0.60 for AMI; r = 0.69 for sepsis; Fig S3G, H), reinforcing the pathological relevance of these modules. Finally, by intersecting the gene constituents of these key modules, we identified 369 overlapping genes, which potentially serve as the core molecular bridge linking the systemic immune dysregulation in sepsis with the acute myocardial injury in AMI (Figure S3I).

3.3. Enrichment analysis of AMI and sepsis co-driver genes

A total of 369 overlapping genes were included in the WGCNA module for AMI and sepsis. In a previous DEG analysis, 53 shared genes were identified between AMI and sepsis. Considering that the module of WGCNA screening contains a set of genes with similar expression profiles, it may not fully cover all the DEGs that are critical for disease progression. Therefore, to avoid missing potentially critical genes, we integrated the 369 WGCNA overlapping genes with the 53 DEG shared genes. After removing the overlapping genes, a total of 417 candidate common driver genes were obtained. These genes may play important roles in common molecular mechanisms of AMI and sepsis. Therefore, GO analysis (Figure1A) and KEGG enrichment analysis (Figure 1B) were first performed on these genes. In order to show the results of KEGG pathway enrichment analysis more clearly, an additional gene dimension was added to the traditional four dimensions (Figure 1C). The above results indicated that these genes were significantly involved in biological processes such as cytokine-mediated signaling pathways, oxidative stress response, and positive regulation of cytokine production and were closely related to pathways such as chemical carcinogenesis and reactive oxygen species.

Figure 1.

Functional enrichment and PPI network analysis. (A,B) GO and KEGG enrichment analysis for 417 candidate driver genes. P-values were adjusted using the Benjamini-Hochberg method, with a significance threshold of FDR < 0.05; (C) The Sankey diagram on the left demonstrates which enrichment pathway the genes belong to, and the bubble diagram on the right demonstrates the gene of the enriched pathway; (D, E) Enrichment analysis of 417 candidate common driver genes by using the Metascape online tool; (F) PPI network analysis; (G) Determination of the top 10 genes by the MCC algorithm; (H) Determination of the top 10 genes by degree ordering; (I) The intersection of genes obtained by the first two algorithms was taken.

In addition, to further elucidate the enrichment pathways of genes related to peripheral circulating markers, analysis of the Metascape database revealed that different genes may exhibit different distributions of functional clusters, with the positive regulation of neutrophil degranulation being the most prominent (Figure 1D). Neutrophil degranulation, as a highly coordinated and robust innate immune response, is the body’s first line of defense against acute infections [13]. Meanwhile, the enrichment analysis of the Metascape database also depicted the common involvement of the core responses of immune defense in the pathogenesis of AMI and sepsis (Figure 1E). Finally, to further screen genes into the same functional group, we entered 417 candidate common driver genes into the STRING database and deleted isolated genes (Figure 1F). Subsequently, the top 10 genes in the PPI network were identified based on the MCC algorithm using Cytoscape software (Figure 1G), and the top 10 genes were further selected based on the degree value sorting (Figure 1H). Taking the intersection of the top 10 genes selected by the two algorithms (Figure 1I), we obtained 7 genes with both high degree and high MCC values, which are most likely to be the most important ‘core hubs’ in the network. Finally, TIMP1, MYD88, SPI1, CD163, FCER1G, ANXA1, and JAK2 were identified as candidate shared hub genes, with MYD88 being the most significant in the group.

3.4. Precision screening of core genes: sharing pivotal genes for diagnostic efficacy assessment based on ROC curves

In this study, TIMP1, MYD88, SPI1, CD163, FCER1G, ANXA1, and JAK2 were identified as shared hub genes for AMI and sepsis. These initial core gene lists were screened based on PPI network topologies (MCC and degree) and represent ‘hubs’ of protein interactions. However, this does not directly reflect their ability to serve as disease diagnostic markers. Therefore, this study used ROC curves to assess the diagnostic predictive value of hub genes in different datasets (Figure 2A, B). In the AMI dataset GSE249812, TIMP1 (AUC = 0.92), MYD88 (AUC = 0.75), JAK2 (AUC = 0.75), FCER1G (AUC = 0.67), SPI1 (AUC = 0.58), ANXA1 (AUC = 0.58), and CD163 (AUC = 0.50) (Figure 2A). This implies that TIMP1 is a very promising single biomarker, as it can achieve a high true positive rate (TPR) while maintaining a very low false positive rate (FPR). The AUC values of MYD88 and JAK2 were both greater than 0.7, suggesting that these two genes also have good discriminatory ability, which is significantly superior to random guessing. FCER1G has some predictive ability, but the application of alone value was not high. SPI1, ANXA1, and CD163, on the other hand, had lower diagnostic efficacy. Also in the sepsis dataset, JAK2 (AUC = 1.00), FCER1G (AUC = 0.99), ANXA1 (AUC = 0.98), MYD88 (AUC = 0.95), TIMP1 (AUC = 0.90), SPI1 (AUC = 0.81), and CD163 (AUC = 0.64) (Figure 2B). Among them, the AUC values of JAK2, FCRER1G, ANXA1, MYD88, TIMP1, and SPI1 were all greater than 0.8, suggesting that these six genes have excellent diagnostic value, whereas CD163 performed weakly. With comprehensive judgment, three genes, TIMP1, MYD88, and JAK2, which had AUC values greater than 0.7 in both datasets, were selected. This suggests that these three genes have good diagnostic performance and can be potential diagnostic markers common to AMI and sepsis. Follow-up studies will explore TIMP1, MYD88, and JAK2 as core shared hub genes in depth.

Figure 2.

ROC identification of hub genes. (A) ROC curves for 7 shared diagnostic markers in the AMI cohort GSE249812; (B) ROC curves for 7 shared diagnostic markers in the sepsis cohort. The Area Under the Curve (AUC) was calculated to assess diagnostic sensitivity and specificity.

3.5. Core gene immune cell infiltration analysis

Since the results of the enrichment analysis indicated that immunity is critical in the development of both diseases, an independent dataset for sepsis was selected for the immune infiltration analysis in this study. This choice was based on the following considerations: Sepsis has an independently validated dataset containing a sufficient sample size and complete clinical information, which ensures the reliability of the results of the immune infiltration analysis. In contrast, the AMI dataset is not sufficient in terms of sample size or sequencing quality for robust immune cell proportion estimation. In our study, we analyzed 24 types of immune cell infiltration using the ImmuCellAI algorithm. The difference in abundance of 19 immune cell types was statistically significant (p < 0.05) between the sepsis group and the NC group. Specifically, the sepsis group had a significantly higher abundance of macrophages (macrophage), neutrophils (neutrophil), and Th1 cells, while the NC group had DC cells, B cells, NK cells, Tgd cells, CD4+ naïve T cells (CD4-naïve), CD8+ naïve T cells (CD8-naïve), and Tr1 cells, and CD4+ T cells, CD8+ T cells, nTreg cells, iTreg cells, Th2 cells, Tfh cells, Tc cells, Tex cells, and MALT cells were in higher abundance (Figure S4A, B). Macrophages and neutrophils, which were significantly enriched in the sepsis group, are classical hallmarks of innate immune activation and acute inflammatory response [14]. Sepsis is triggered by an infection, and the organism first mobilizes these cells to phagocytose and clear pathogens. Their significant enrichment is fully consistent with the pathophysiology of sepsis and is the cellular basis of the ‘inflammatory storm.’ Th1 cells primarily secrete cytokines, such as IFN-γ, that drive the cellular immune response against intracellular pathogens (e.g. viruses and some bacteria) [15]. Their enrichment suggests that the adaptive immune system is also strongly activated and is favoring Th1-type immunity. favors a Th1-type immune response. Taken together, the immune profile of the sepsis group showed strong innate immune inflammation and a Th1-type adaptive immune response. To deepen our understanding of the functional importance of 3 key genes in immune cell infiltration, Spearman correlation analysis was performed in this study to examine the association between the expression levels of these genes and immune cell abundance. The results showed that the three core genes, TIMP1, MYD88, and JAK2, all showed strong positive correlations with the abundance of macrophages (Macrophage) and neutrophils (Neutrophil) (Figure S4C). This again demonstrates the involvement of these characterized genes in the immune cell infiltration in the blood immune microenvironment of sepsis patients, thereby influencing disease progression.

3.6. Single-cell analysis of hub gene locations

The number of high-quality cells was distributed in 2716–9485 for each sample after quantitative quality control by Cell Ranger. The final number of cells obtained was distributed in 2197–8676 after the quality control of eliminating double cells, multiple cells, and apoptotic cells. The average number of UMIs per cell was distributed in 4194–7859, the average number of genes was distributed in 1173–1909, and the average percentage of mitochondrial UMIs was distributed in 0.0600–0.1100. A total of 22 cell clusters were classified into 22 cell clusters after dimensionality reduction clustering (Figure S5 A), and the cells were further categorized into 6 major cell clusters (Figure S5 B), including neutrophils (Neutrophils), NK cells (NK_cells), platelets (Platelets), B cells (B_cells), monocytes (Monocytes), and T cells (T_cells). To bridge the gap between single-cell and bulk transcriptomics, we performed differential expression analysis within each cluster. In the monocyte population, genes associated with inflammatory response and myeloid activation were significantly enriched. Functional enrichment analysis via Gene Set Variation Analysis (GSVA) revealed that these clusters were predominantly involved in the Toll-like receptor signaling pathway, TNF-alpha signaling via NF-kB, and IL-6/JAK/STAT3 signaling, which aligns with the key mechanistic pathways identified in our previous bulk RNA-seq analysis.

Subsequently, we performed localization analysis of key diagnostic markers and found that most markers were commonly expressed in monocytes (Figure S5C). The cell clustering results were validated by analyzing the expression of key marker genes (Figure S5D). All three genes were highly expressed in Clusters 8 and 16, i.e. in monocyte clusters. Importantly, the expression trends of these hub genes in the single-cell dataset—showing significant up-regulation in the monocyte clusters of sepsis patients compared to healthy controls—were highly consistent with the expression patterns observed in the GSE249812 and clinical validation cohorts. These findings strongly validate our cell clustering strategy and underscore the critical role of monocyte-mediated immune dysregulation in the pathogenesis of both AMI and sepsis.

3.7. Prognostic Meta-analysis of core genes in sepsis

To preliminarily validate the prognostic potential of the core genes JAK2, MYD88, and TIMP1 in a wide range of sepsis patients and to assess the robustness and generalizability of the results, this study performed a meta-analysis of transcriptomic and clinical data from multiple independent cohorts (GSE163151, GSE232753, GSE95233, GSE69063, GSE243217, GSE236713, GSE185263, GSE154918, GSE134347, and GSE100159). The results showed that the expression of these core genes was significantly higher in the sepsis group than in the NC group. As shown in Figure 3A–D, this expression difference was statistically significant (p < 0.05). This greatly enhances the reliability of the findings of this study and provides solid evidence for its future application to clinical prognosis determination. Notably, utilizing the sepsis-derived single-cell landscape to infer the mechanisms of AMI is grounded in the ‘Common Inflammatory Pathway’ hypothesis. Since both sepsis and AMI share a core pathological axis driven by monocytic dysregulation and intense systemic inflammation, this high-resolution sepsis model serves as a robust systemic framework for identifying conserved regulatory hubs and myeloid cell evolution, thereby providing critical insights into the sterile inflammatory mechanisms of AMI.

Figure 3.

Meta-analysis of core hub genes. (A-C) Forest plots representing JAK2, MYD88, and TIMP1 expression across 10 GEO datasets. A random-effects or fixed-effects model was applied based on the I² heterogeneity index. (D) Expression levels of core genes in the NC and sepsis groups. Data are presented as box plots, and statistical significance was determined using the MultiMeta algorithm (p < 0.05).

3.8. Hub gene-based drug candidate screening and molecular docking

Based on the DSigDB library in the Enrichr platform, drug candidates with significant corrected P-values were screened by calculating p-values and binding scores to core hub genes. These potential small molecule compounds are expected to serve as combination therapeutic agents for AMI and sepsis (Table 2). By combining the p-value, binding score, and association with the core hub genes TIMP1, JAK2, and MYD88, the drug candidate was fine-screened to quercetin (Quercetin). This implies that quercetin may be more effective in ‘reversing’ disease gene expression profiles. Quercetin is a natural flavonoid widely found in food, with a high safety profile and a low threshold for research and use [16]. In order to verify the feasibility of these computational predictions at the structural level and to elucidate its potential molecular mechanism of action, the present study conducted molecular docking simulations of quercetin with the above three pivotal gene targets, JAK2, MYD88, and TIMP1. To investigate the interaction of small molecule active ingredients with protein targets, molecular docking was performed in this study using MOE 2019 software. According to the docking results, the molecular docking energies ranged from −5.8099 to −6.5539 kcal/mol. It is generally accepted that a docking energy value of less than −4.25 kcal/mol indicates the presence of some binding activity between the two, that less than −5.0 kcal/mol indicates good binding activity, and that less than −7.0 kcal/mol indicates strong binding activity. The docking results showed that quercetin had good binding activity to all three targets (Figure 4A–C). Among them, the binding of quercetin to JAK2 (−6.5539 kcal/mol) showed lower docking binding energy. The lower binding energy implies that the small molecule active ingredient may have stronger binding activity to the target site.

Table 2.

Identification of drug candidates based on key genes.

Term	P-value	Adjusted P-value	Combined Score	Genes
quercetinCTD00006679	0.0039336	0.019647068	279822.8361	TIMP1;JAK2;MYD88
WP1066 CTD 00004704	0.0016491	0.019647068	6403.334808	JAK2
meclizineCTD00006252	0.0016491	0.019647068	6403.334808	TIMP1
MevastatinTTD00009287	0.0016491	0.019647068	6403.334808	JAK2
OxibendazoleCTD00000598	0.001799	0.019647068	5741.921288	JAK2
Cladribine TTD 00007218	0.001799	0.019647068	5741.921288	JAK2
Erlotinib TTD 00007882	0.001799	0.019647068	5741.921288	JAK2
calcitriol TTD 00002649	0.001799	0.019647068	5741.921288	JAK2
Trifluridine TTD 00011552	0.001799	0.019647068	5741.921288	JAK2
AZARIBINE CTD 00003341	0.0019488	0.019647068	5196.553512	JAK2
AM-630 CTD 00003207	0.0019488	0.019647068	5196.553512	TIMP1
methazolamideCTD00006289	0.0019488	0.019647068	5196.553512	TIMP1
herbimycin a CTD00001010	0.0019488	0.019647068	5196.553512	TIMP1
CHEMBL55802CTD00003118	0.0022484	0.019647068	4351.685757	TIMP1
TITANIUM CTD 00006899	0.0022484	0.019647068	4351.685757	TIMP1

The top 15 drug candidates were screened based on the binding energy sorted from largest to smallest, all with statistically significant differences at a p-value < 0.05.

Figure 4.

Molecular docking of Quercetin with hub proteins. (A-C) 3D and 2D interaction maps of Quercetin with JAK2, MYD88, and TIMP1 protein receptors. Hydrogen bonds, hydrophobic interactions, and electrostatic interactions are highlighted. Binding affinities were quantified by the lowest binding energy (kcal/mol).

3.9. In vitro experimental verification of core genes and targeted drugs

To investigate the roles of the three core genes identified: JAK2, MYD88, and TIMP1 as well as the targeted drug Quercetin in sepsis and myocardial infarction, we constructed two cell models. Specifically, we utilized LPS induced RAW264.7 cells to establish a sepsis model and developed a H/R model to simulate myocardial infarction. The H/R model was validated through cell viability assays (Figure 5A). Quantitative PCR analysis revealed that the mRNA levels of the inflammatory factors IL-6 and TNF-α were significantly elevated in both the sepsis and myocardial infarction cell models (Figure 5B,C). Subsequently, we examined the mRNA expression of the core genes. The results indicated that the expression levels of JAK2, MYD88, and TIMP1 aligned with our differential gene expression analysis, revealing significant elevations in mRNA levels in both disease cell models. Consequently, we confirmed that JAK2, MYD88, and TIMP1 are critical genes involved in the progression of sepsis and myocardial infarction. Furthermore, the mRNA expression levels of inflammatory factors and these key genes were markedly reduced in both cell models following the administration of Quercetin, the targeted drug analyzed above (Figure 5B–E).

Figure 5.

In vitro experimental verification. (A) CCK-8 assay for H9c2 cell viability; (B-E) qPCR validation of inflammatory cytokines (TNF-α, IL-6) and hub genes (Myd88, Jak2, Timp1) in RAW264.7 and H9c2 cells. All data are expressed as mean ± SD from at least three independent experiments. Statistical analysis was performed using one-way ANOVA followed by Tukey’s post-hoc test for multiple groups, and Student’s t-test for comparisons between two groups. Significance levels: *p < 0.05, **p < 0.01, ***p < 0.001.

4. Discussion

The incidence of AMI is significantly correlated with an increased susceptibility to sepsis, characterized by an exacerbated inflammatory state. Patients presenting with concurrent AMI and sepsis face a markedly poorer prognosis compared to those with sepsis alone [17]. Despite their distinct clinical etiologies, both conditions converge on shared immune-inflammatory responses, cellular apoptosis, and patho­logical tissue damage. Investigating these shared genetic drivers reveals the underlying molecular pathways—such as uncontrolled inflammatory cascades and leukocyte dysregulation—that propel disease progression in both scenarios.

Our study identified JAK2, MYD88, and TIMP1 as core genes, offering a molecular basis for precision risk stratification. These findings facilitate the identification of high-risk patient subgroups, specifically those with coronary artery disease who possess an elevated genetic predisposition to lethal hyper-inflammation during secondary infections. By integrating these biomarkers into multidimensional predictive models, clinicians can achieve earlier clinical warnings and more targeted patient screening. For instance, a validated machine learning model utilizing such clinical and molecular data has already demonstrated robust capabilities in detecting AMI within septic populations, enabling proactive rather than reactive intervention [18].

JAK2: As a mediator of the IL-6-driven hyper-inflammatory surge in early sepsis, JAK2 also participates in compensatory anti-inflammatory responses, exhibiting a complex dual nature [19]. Similarly, in AMI, JAK2 activity rises rapidly to facilitate inflammatory cell infiltration and cardiomyocyte survival [20]. This duality suggests that JAK2 inhibition is a ‘double-edged sword’; its clinical application requires precise temporal gating to mitigate early cytokine storms without hindering essential post-infarction cardiac repair.

MYD88: Serving as the central hub for innate immune activation, MYD88 drives the initial excessive inflammatory response in sepsis [21,22]. In AMI, it orchestrates the sterile inflammation triggered by endogenous danger signals from necrotic tissue [23]. Targeting the MYD88 pathway offers a window for early-phase immunomodulation to prevent collateral tissue damage.

TIMP1: This matrix metalloproteinase inhibitor serves as a potent prognostic indicator. Beyond its role in plaque stabilization and cardiac fibrosis [24], TIMP1 regulates macrophage polarization and lymphocyte apoptosis. Its capacity to predict mortality in critical sepsis [25] and its association with sepsis susceptibility [26] underscore its value as a dynamic biomarker for monitoring immune exhaustion.

Quercetin, a natural flavonoid, demonstrates significant therapeutic potential in both sepsis and AMI by modulating these very pathways—reducing infarct size, improving organ function, and enhancing survival in preclinical models [27–29]. However, translating these findings requires optimizing bioavailability and conducting rigorous trials to determine the optimal therapeutic window.

Sepsis and AMI are initiated by microbial invasion and atherosclerotic thrombosis, respectively [30]. Despite these divergent etiologies, both conditions converge on shared cellular and molecular pathways during mid-to-late disease progression. While investigating these commonalities is scientifically pertinent, caution is warranted, as an overemphasis on shared mechanisms must not obscure the unique initiating events of each condition. Pathophysiologically, these disorders create a complex network of bidirectional interactions. Sepsis-induced systemic inflammation and hemodynamic instability can directly compromise myocardial function [31,32], while the robust systemic inflammatory response following a massive myocardial infarction may manifest a sepsis-like phenotype [33]. A central challenge in investigating this shared genetic basis is distinguishing primary pathogenic drivers from secondary phenotypic effects. For instance, while suppressing early-stage inflammation in sepsis may mitigate tissue damage, an identical intervention during the AMI repair phase could hinder macrophage-mediated clearance and restructuring, potentially leading to adverse ventricular remodeling [34]. This underscores the critical importance of spatiotemporal specificity and dosage in pharmacological strategies.

Beyond shared immuno-inflammatory pathways, sepsis and AMI exhibit significant overlap in clinical risk factors and organ dysfunction patterns, which profoundly influence prognosis. Elevated serum lactate—reflecting systemic tissue hypoperfusion and cellular metabolic derangement—along with acute kidney injury (AKI) or pre-existing renal dysfunction, serve as potent predictors of mortality in both critically ill sepsis patients and ST-segment elevation myocardial infarction (STEMI) patients complicated by cardiogenic shock (CS) [35,36]. Specifically, renal impairment significantly impacts long-term survival in STEMI-CS patients undergoing primary percutaneous coronary intervention (PCI), mirroring the prognostic weight of AKI in septic shock [37].

Furthermore, the integration of laboratory and clinical parameters into global risk scores, such as the Intermountain Risk Score (IMRS), has proven more effective at predicting short- and long-term outcomes in CS and sepsis patients than single biomarkers alone [38]. Incorporating our identified core genes—JAK2, MYD88, and TIMP1—into such multidimensional risk frameworks (comprising lactate levels, renal function, and validated clinical scores like IMRS) could significantly enhance clinical decision-making.

Given the current scarcity of AMI gene expression cohorts with comprehensive survival data, this study validated the clinical significance of these core genes using a sepsis cohort. However, our in vitro experiments confirmed the functional roles of these genes in both diseases, and the targeted agent quercetin demonstrated potential therapeutic efficacy. Future research should prioritize direct clinical validation of these genetic signatures and their synergy with clinical risk parameters in the specific context of acute myocardial infarction.

5. Conclusions

In this study, we successfully identified JAK2, MYD88, and TIMP1 as core genes shared between sepsis and acute myocardial infarction. These genes not only serve as potential diagnostic biomarkers but also offer new targets for the development of common therapeutic strategies for these two diseases, with quercetin predicted as a potential therapeutic candidate.

Funding Statement

This work was supported by Luzhou Municipal People’s Government-Southwest Medical University Science and Technology Strategic Cooperation Project [00140256, 2024LZXNYDJ115]; Sichuan Provincial Collaborative Innovation Center for Cardiovascular Disease Prevention and Treatment Fund Project [00160082, No.XTCX2019-14].

Ethics approval

The studies involving humans were approved by The Ethical Committee of the Affiliated Hospital of Southwest Medical University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data availability Participant data uploaded. RNA-seq data for 23 sepsis patients and 10 healthy volunteers are available in the China National Genome Database (CNGBdb) using the following link: https://db.cngb.org/, Accession Code: CNP0002611. The data that support the findings of this study are available from the corresponding author upon reasonable request.