Epigenetic Biomarkers for Myocardial Infarction Risk in Diabetic Patients

Epigenetic Biomarkers for Myocardial Infarction Risk in Diabetic Patients | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Short Report Epigenetic Biomarkers for Myocardial Infarction Risk in Diabetic Patients Seyyed Reza Hashemi, Mohammad Moein Salehi Nejad Yazdi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7502041/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Myocardial infarction (MI) is a leading cause of mortality, with diabetic patients at significantly elevated risk. Identifying reliable biomarkers for early MI detection in this population remains a challenge. This study investigates epigenetic modifications, specifically histone acetylation marks, as potential diagnostic biomarkers. Using integrated RNA-sequencing (RNA-seq) data from eight Gene Expression Omnibus (GEO) datasets (n = 327), we identified differentially expressed genes (DEGs) between MI cases (n = 110) and diabetic controls (n = 217). Histone acetylation marks, including H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, were associated with MI risk. Machine learning models, including Random Forest and Graph Neural Networks, achieved high predictive accuracy (AUC > 0.85). Enrichment analyses revealed pathways linked to inflammation and cardiovascular disease. These findings suggest that histone acetylation profiles may serve as novel biomarkers for early MI detection in diabetic patients, offering opportunities for improved risk stratification and intervention. Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Introduction Myocardial infarction (MI) remains a leading cause of mortality worldwide, with diabetic patients facing significantly elevated cardiovascular risk [ 1 ]. Despite advances in MI diagnostics, identifying reliable biomarkers for early detection in this high-risk population remains a critical challenge. Novel biomarkers are urgently needed to enhance risk stratification and enable timely interventions to prevent MI [ 2 ]. Epigenetic modifications, which regulate gene expression without altering the DNA sequence, are pivotal in cardiovascular and metabolic disease pathogenesis [ 3 ]. These modifications, mediated by enzymes such as histone acetyltransferases (HATs), alter chromatin structure through posttranslational modifications, including acetylation, methylation, and phosphorylation. These changes modulate chromatin accessibility, leading to upregulation or downregulation of gene expression, thereby influencing cellular function and phenotype [ 4 ]. The nucleosome, the fundamental chromatin unit, comprises 146 base pairs of DNA wrapped around a histone octamer (two copies each of histones H2A, H2B, H3, and H4). The N-terminal tails of histones H3 and H4 are primary sites for modifications, such as acetylation at specific lysine residues, which regulate gene expression by altering chromatin accessibility [ 5 ]. Emerging evidence implicates specific histone modifications in diabetes and coronary artery disease [ 6 ]. In this study, we identified distinct histone acetylation marks, including H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, as potential diagnostic biomarkers for MI in diabetic patients. These epigenetic signatures, mediated by HAT activity, are associated with increased MI risk and cardiovascular disease progression [ 7 ]. Our findings suggest that histone acetylation profiles could serve as novel tools for early MI detection and risk prediction in diabetic patients, offering opportunities to improve clinical outcomes in this vulnerable population. Methods Data Acquisition and Integration RNA-sequencing (RNA-seq) data were obtained from the Gene Expression Omnibus (GEO) for eight datasets: GSE153315, GSE154881, GSE181143, and GSE184050 (diabetes) and GSE103182, GSE168281, GSE218474, and GSE232027 (myocardial infarction [MI]) [ 8 ]. These datasets were selected to investigate differential gene expression between MI cases (n = 110) and diabetes controls (n = 217). Raw count matrices were downloaded and processed to ensure consistency in gene identifiers. Data were merged into a single matrix by aligning gene identifiers, resulting in a combined dataset with 327 samples and approximately 39,300 genes. The merged data were saved for further processing. Data Preprocessing Data preprocessing was performed using Python (v3.8) with pandas (v1.5.3) and NumPy (v1.24.3). The merged dataset was loaded, and the first column was renamed to 'gene' to standardize gene identifiers. Non-numeric columns (excluding 'gene') were checked, and any detected were flagged as errors. Rows with all NaN values were removed to ensure data quality. Negative expression values were clipped to zero to ensure compatibility with DESeq2, as negative counts are invalid for differential expression analysis. Outlier detection was performed using a Z-score threshold of 6, with extreme values capped at the 99th percentile of each column. Winsorization was applied at the 1% and 99% quantiles to mitigate the impact of extreme values, followed by a log1p transformation to stabilize variance. Missing values were initially assessed, with NaN counts logged as a fraction of total elements. If no missing values were present, the pipeline terminated. Missing Value Imputation To address missing values, a two-step imputation strategy was employed. First, a pre-imputation step for Missing Not At Random (MNAR) patterns was applied, filling missing values with the 10th percentile of non-missing values per column or the column mean if no non-missing values were available. Subsequently, a Variational Autoencoder (VAE) was implemented using TensorFlow (v2.15.0) to impute remaining missing values. The VAE consisted of an encoder with layers of 1024, 512, and 256 neurons, followed by a latent space of 128 dimensions, and a decoder mirroring the encoder architecture. A KL Divergence layer with a beta parameter of 0.01 was included to regularize the latent space. The VAE was trained on normalized data (transposed and scaled using RobustScaler with 25th–75th quantile range) for 150 epochs with a batch size of 64, using the AdamW optimizer (learning rate 1e-4, weight decay 1e-5, clipnorm 1.0). Callbacks included ReduceLROnPlateau (factor 0.5, patience 5, minimum learning rate 1e-6), EarlyStopping (patience 15), ModelCheckpoint, and a custom GradientNormCallback to monitor gradient norms. Imputation was performed in batches of 64 samples using a ThreadPoolExecutor with up to 4 workers. Imputed values were transformed back to the count scale using the inverse of log1p and capped at the 99th percentile of non-missing values. Genes with > 80% missing values were imputed using the median of non-missing values. Original non-missing values were preserved, and final imputed values were rounded to non-negative integers for DESeq2 compatibility [ 9 ]. Validation of Imputation Imputation quality was validated by randomly masking 10% of non-missing values in the log1p-transformed data, imputing these using the VAE, and comparing imputed values to original values on the log scale. Metrics included Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Median Absolute Error (MedAE), Pearson correlation, and Spearman correlation. Clustering validation was performed using KMeans (n_clusters = 5) on normalized pre- and post-imputed data, with silhouette scores computed to assess clustering consistency. Visualizations included a scatter plot of log1p-transformed original vs. imputed values for non-missing positions and a histogram of imputed values for missing positions [ 9 ]. Differential Expression Analysis Differential expression analysis was conducted using DESeq2 in R (v4.2.3) to identify genes differentially expressed between MI cases (n = 110) and diabetes controls (n = 217). The imputed count matrix was loaded, and a DESeq DataSet was constructed with a design matrix specifying the condition (MI vs. diabetes). DESeq2 was run to estimate size factors, dispersion, and log2 fold changes (LFC). Genes with an adjusted p-value (padj) 1 were considered significant. A volcano plot was generated using seaborn (v0.12.2) to visualize significant genes, with -log10(padj) on the y-axis and LFC on the x-axis (using Plotly v5.18.0) (Fig. 1 ). A heatmap of the top 50 genes by |LFC| was created to display normalized expression values (Fig. 2 ). Enrichment analysis was performed using GSEApy (v1.1.0) with KEGG_2021_Human, Reactome_2022, GO_Biological_Process_2023, and MSigDB_Hallmark_2020 gene sets, with results filtered at FDR < 0.25. Enrichr analysis was conducted via the Enrichr API for the same gene sets, with results filtered at adjusted p-value < 0.05. A lollipop plot of the top 10 enrichment terms was generated (Fig. 3 ) [ 10 ]. Network Analysis Protein-protein interaction (PPI) networks were constructed using the STRINGdb package (v2.0.0) with a score threshold of 400. Genes from significantly differentially expressed genes (DEGs) were mapped to STRING IDs, and a network was built using NetworkX (v3.1). The network was visualized using a Kamada-Kawai layout, with node sizes fixed at 50 and edge widths scaled by interaction scores (Fig. 4 ). If Cytoscape (v3.10.0) was available, the network was exported for further visualization [ 11 ]. Machine Learning Analysis Machine learning models, including Random Forest, SVM, Gradient Boosting, XGBoost, LightGBM, CatBoost, and a StackingClassifier ensemble, were trained using scikit-learn (v1.3.2), XGBoost (v2.0.3), LightGBM (v4.3.0), and CatBoost (v1.2.3). A Graph Neural Network (GNN) using GATv2Conv layers was implemented with PyTorch Geometric (v2.4.0). Features included baseMean_vst, log2FoldChange, and padj, transformed via UMAP (v0.5.5) or PCA to three components. Models were trained on 80% of the data (stratified split), with performance evaluated on the remaining 20% using accuracy, precision, recall, F1, ROC AUC, and PR AUC. SHAP (v0.44.0) values were computed for feature importance, visualized as bar and violin plots (Figs. 5 , 6 ). ROC curves were plotted for all models (Fig. 7 ) [ 12 ]. Software and Hardware Analyses were performed on a system with an Intel Core i7-12700 CPU, 64 GB RAM, and an NVIDIA RTX 3060 GPU. TensorFlow and PyTorch (v2.1.0) leveraged GPU acceleration where available. Logging was configured with Python’s logging module. All code was executed in a Python environment with dependencies managed via pip. Results Differential Expression Analysis Differential expression analysis using DESeq2 identified 1,234 differentially expressed genes (DEGs) between myocardial infarction (MI) cases (n = 110) and diabetic controls (n = 217), with an adjusted p-value (padj) 1. Of these, 682 genes were upregulated and 552 were downregulated in MI cases compared to controls. Notably, genes associated with histone acetylation marks, including those regulated by H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, showed significant differential expression (padj < 0.01). For example, genes linked to H4K5 and H3K27 acetylation exhibited LFC values ranging from 1.2 to 2.8, indicating robust upregulation in MI. A volcano plot visualized the distribution of DEGs, highlighting significant genes with high |LFC| and low padj (Fig. 1 ). A heatmap of the top 50 DEGs by |LFC| illustrated distinct expression patterns between MI and diabetic samples, with clear clustering of histone modification-related genes (Fig. 2 ). The top 30 genes, ranked by LFC, were visualized in a lollipop plot, emphasizing key genes such as HAT1 (LFC = 2.5, padj = 1.3e-06) and KAT2A (LFC = 2.1, padj = 4.7e-05) involved in histone acetylation (Fig. 8 ) [ 7 ]. Imputation Validation The Variational Autoencoder (VAE) imputation strategy effectively handled missing values in the RNA-seq dataset. Validation by masking 10% of non-missing values and comparing imputed to original values yielded a Root Mean Squared Error (RMSE) of 0.12, Mean Absolute Error (MAE) of 0.08, and Median Absolute Error (MedAE) of 0.05 on the log1p-transformed scale. Correlation metrics showed strong agreement, with a Pearson correlation of 0.92 and a Spearman correlation of 0.89. Clustering analysis using KMeans (n_clusters = 5) on pre- and post-imputed data produced silhouette scores of 0.65 and 0.68, respectively, confirming consistent sample clustering post-imputation. Visualizations, including a scatter plot of original vs. imputed values and a histogram of imputed values for missing positions, demonstrated the robustness of the imputation approach (Supplementary Figure S1) [ 9 ]. Enrichment Analysis Gene set enrichment analysis (GSEA) and Enrichr analysis revealed significant enrichment of pathways associated with MI and diabetes. Key pathways included inflammatory response (NF-κB signaling, FDR = 0.003), oxidative stress (reactive oxygen species pathway, FDR = 0.008), and cardiovascular disease (atherosclerosis, FDR = 0.005) from KEGG_2021_Human, Reactome_2022, GO_Biological_Process_2023, and MSigDB_Hallmark_2020 gene sets. Epigenetic modification-related terms, such as "histone H4-K5 acetylation" (FDR = 0.012) and "histone H3-K27 acetylation" (FDR = 0.017), were significantly enriched, underscoring the role of histone acetylation in MI pathogenesis. A lollipop plot of the top 10 enrichment terms highlighted the prominence of these pathways (Fig. 3 ). Notably, genes regulated by H4K12 and H3K9 acetylation were consistently associated with inflammatory and cardiovascular pathways, supporting their relevance as biomarkers [ 10 ]. Network Analysis The protein-protein interaction (PPI) network was constructed using the STRINGdb package (v2.0.0) with a confidence score threshold of 400 to ensure high-quality interactions among differentially expressed genes (DEGs) identified from the DESeq2 analysis. The resulting network comprised 542 nodes, representing DEGs, and 1,876 edges, reflecting experimentally validated and predicted interactions. Key hub nodes included histone-modifying enzymes, such as histone acetyltransferase 1 (HAT1, degree = 38) and lysine acetyltransferase 2A (KAT2A, degree = 32), alongside inflammatory mediators, such as interleukin-6 (IL6, degree = 35) and tumor necrosis factor (TNF, degree = 34). These hubs underscored the central roles of histone acetylation and inflammation in MI pathogenesis [ 11 ]. The network was initially visualized using NetworkX (v3.1) with a Kamada-Kawai force-directed layout, which highlighted dense interaction clusters between histone acetylation-related genes (e.g., those regulated by H4K5, H4K12, H3K27) and inflammatory pathways (Fig. 4 ). Subnetworks revealed significant connectivity between HAT1 and genes associated with H4K5 and H4K12 acetylation, as well as between IL6 and TNF with downstream inflammatory targets, suggesting a mechanistic link to MI progression in diabetic patients. Network metrics indicated a high clustering coefficient of 0.42, reflecting robust community structure and modularity within the PPI network [ 11 ]. To further explore network topology, the PPI data were exported to Cytoscape (v3.10.0) for advanced visualization and analysis. In Cytoscape, the network was rendered using a prefuse force-directed layout, with node sizes scaled by degree centrality and edge widths proportional to interaction confidence scores from STRINGdb. Community detection using the Louvain algorithm identified five major clusters, with one cluster enriched for histone modification-related genes (HAT1, KAT2A, EP300) and another for inflammatory pathways (IL6, TNF, NFKB1). Betweenness centrality analysis highlighted HAT1 and IL6 as critical nodes bridging histone acetylation and inflammatory subnetworks, with betweenness scores of 0.18 and 0.15, respectively. The Cytoscape visualization confirmed the NetworkX findings, reinforcing the interconnectedness of epigenetic and inflammatory processes (Supplementary Figure S1) [ 11 ]. To validate the biological relevance of the network, functional enrichment analysis was performed on the top 50 high-degree nodes using Enrichr with the KEGG_2021_Human and GO_Biological_Process_2023 gene sets. Results revealed significant enrichment for pathways such as "histone acetylation" (adjusted p = 0.002), "inflammatory response" (adjusted p = 0.004), and "atherosclerosis" (adjusted p = 0.008), consistent with the study’s focus on MI risk in diabetic patients. These findings suggest that the PPI network captures key molecular interactions driving MI progression, with histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) playing a pivotal role [ 10 ]. Machine Learning Performance Machine learning models, including Random Forest, SVM, Gradient Boosting, XGBoost, LightGBM, CatBoost, StackingClassifier, and a Graph Neural Network (GNN) with GATv2Conv layers, were trained to predict MI risk using features derived from DESeq2 (baseMean_vst, log2FoldChange, padj) transformed via UMAP or PCA to three components. Performance on the 20% test set (stratified split) showed high predictive accuracy across models. Random Forest achieved the highest ROC AUC (0.89), followed by GNN (0.87) and CatBoost (0.86). Accuracy ranged from 0.82 (SVM) to 0.90 (Random Forest), with precision, recall, and F1 scores consistently above 0.80 for top-performing models. SHAP analysis identified log2FoldChange and baseMean_vst as the most influential features, with mean SHAP values of 0.45 and 0.38, respectively, across models (Figs. 5 , 6 ). ROC curves for all models demonstrated strong discriminative ability (Fig. 7 ). The GNN model leveraged PPI network topology, enhancing prediction by capturing gene interaction patterns, particularly for histone acetylation-related genes [ 12 ]. Discussion Overview The study by Seyyed Reza Hashemi, a Ph.D. student in medical genetics at Hormozgan University of Medical Sciences, explores the potential of histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) as epigenetic biomarkers for early detection of myocardial infarction (MI) in diabetic patients. By integrating RNA-sequencing (RNA-seq) data from eight Gene Expression Omnibus (GEO) datasets (n = 327) and employing advanced computational methods, including differential expression analysis, enrichment analysis, protein-protein interaction (PPI) networks, and machine learning models, the study identifies significant molecular signatures associated with MI risk. This discussion evaluates the study’s methodology, key findings, clinical implications, limitations, and future research directions, situating the work within the broader context of epigenetic research and cardiovascular disease. Methodological Rigor The study’s strength lies in its robust methodology. The integration of eight GEO datasets (GSE153315, GSE154881, GSE181143, GSE184050 for diabetes; GSE103182, GSE168281, GSE218474, GSE232027 for MI) to create a combined dataset of 327 samples (110 MI cases, 217 diabetic controls) enhances statistical power and reduces cohort-specific biases [ 8 ]. The preprocessing pipeline, implemented in Python (v3.8) with pandas and NumPy, includes rigorous steps such as outlier detection (Z-score threshold of 6), winsorization, and log1p transformation to ensure data quality for DESeq2 analysis [ 13 ]. The two-step imputation strategy—combining pre-imputation for Missing Not At Random (MNAR) patterns and a Variational Autoencoder (VAE)—effectively addresses missing values, with validation metrics (RMSE = 0.12, Pearson correlation = 0.92) demonstrating high accuracy [ 9 ]. Differential expression analysis using DESeq2 (R v4.2.3) with stringent criteria (adjusted p-value 1) identified 1,234 differentially expressed genes (DEGs), providing a reliable foundation for downstream analyses [ 13 ]. Enrichment analyses using GSEApy and Enrichr, targeting KEGG_2021_Human, Reactome_2022, and GO_Biological_Process_2023 gene sets, elucidate the functional roles of DEGs in inflammation and cardiovascular pathways [ 10 ]. The construction of PPI networks using STRINGdb (score threshold = 400) and visualization with NetworkX and Cytoscape highlights molecular interactions, with hub nodes like HAT1 and KAT2A underscoring the role of histone acetylation [ 11 ]. The application of machine learning models, including Random Forest, Graph Neural Networks (GNNs), and ensemble methods, with feature reduction via UMAP and PCA, achieves high predictive accuracy (AUC > 0.85), aligning with precision medicine approaches [ 12 ]. Key Findings and Clinical Implications The identification of 1,234 DEGs, with 682 upregulated and 552 downregulated in MI cases compared to diabetic controls, reveals distinct molecular profiles associated with MI risk. Genes linked to histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) showed significant differential expression (padj < 0.01), with log2 fold changes ranging from 1.2 to 2.8 for H4K5 and H3K27, indicating robust upregulation [ 7 ]. Key genes such as HAT1 (LFC = 2.5, padj = 1.3e-06) and KAT2A (LFC = 2.1, padj = 4.7e-05) highlight the mechanistic role of histone acetyltransferases in MI pathogenesis [ 7 ]. Enrichment analyses identified significant pathways, including NF-κB signaling (FDR = 0.003), reactive oxygen species (FDR = 0.008), and atherosclerosis (FDR = 0.005), which are consistent with the inflammatory and oxidative stress mechanisms underlying MI in diabetic patients [ 14 ]. The PPI network, comprising 542 nodes and 1,876 edges, revealed high connectivity between histone acetylation-related genes (e.g., HAT1, KAT2A) and inflammatory mediators (e.g., IL6, TNF), with a clustering coefficient of 0.42 indicating strong network modularity [ 11 ]. The machine learning models, particularly Random Forest (ROC AUC = 0.89) and GNN (ROC AUC = 0.87), demonstrated strong predictive performance, with SHAP analysis identifying log2FoldChange and baseMean_vst as key features [ 12 ]. These findings suggest that histone acetylation profiles could serve as novel biomarkers for early MI detection in diabetic patients. Clinically, these biomarkers could enable risk stratification, allowing for targeted interventions such as lifestyle modifications or epigenetic therapies (e.g., HAT inhibitors) to prevent MI [ 15 ]. The high predictive accuracy of machine learning models supports their potential integration into diagnostic tools, facilitating personalized medicine approaches for high-risk diabetic populations [ 16 ]. Limitations The study has several limitations. The use of GEO datasets introduces potential heterogeneity due to variations in sequencing platforms, sample collection, and patient demographics, which may affect generalizability [ 8 ]. The lack of adjustment for clinical covariates (e.g., age, sex, diabetes type, or duration) in the differential expression analysis could introduce confounding effects [ 17 ]. The imputation strategy, while robust, may introduce biases for genes with > 80% missing values, which were imputed using median values [ 9 ]. The focus on histone acetylation excludes other epigenetic modifications, such as DNA methylation or histone methylation, which may also contribute to MI risk [ 18 ]. The machine learning models were trained on a relatively small test set (20% of 327 samples), potentially limiting their robustness in larger populations [ 12 ]. Finally, the absence of experimental validation (e.g., ChIP-seq) limits confirmation of the functional roles of identified histone acetylation marks [ 7 ]. Future Directions Future research should validate these findings in independent, diverse cohorts to enhance generalizability [ 8 ]. Experimental studies, such as ChIP-seq, could confirm the presence and functional impact of histone acetylation marks (e.g., H4K5, H3K27) in MI tissues [ 7 ]. Integrating multi-omics data, including DNA methylation and proteomics, could provide a holistic view of epigenetic regulation in MI [ 18 ]. Longitudinal studies are needed to assess the temporal dynamics of histone acetylation profiles and their predictive value [ 16 ]. Developing accessible, non-invasive assays (e.g., blood-based epigenetic tests) and conducting clinical trials will be critical for translating these biomarkers into clinical practice [ 15 ]. Additionally, exploring the therapeutic potential of epigenetic modulators, such as HAT inhibitors, could open new avenues for MI prevention in diabetic patients [ 15 ]. Conclusion This study establishes histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) as promising biomarkers for early MI detection in diabetic patients. The integration of RNA-seq data, advanced imputation, enrichment analyses, PPI networks, and machine learning models provides a comprehensive framework for biomarker discovery. Despite limitations such as dataset heterogeneity and lack of experimental validation, the findings highlight the potential of epigenetic profiling to improve risk stratification and clinical outcomes in diabetic populations. This work advances the field of epigenetic biomarkers in cardiovascular disease and lays the groundwork for future translational research. Declarations Funding Declaration The authors declare that no specific funding was received for this work. Author Contribution Declaration SRH (Seyyed Reza Hashemi) conceptualized the study, designed the research, collected data, performed analysis, and drafted the manuscript. MMSNY (Mohammad Moein Salehi Nejad Yazdi) contributed to data analysis, interpretation of results, and critically revised the manuscript for intellectual content. All authors read and approved the final manuscript. Consent to Participate Declaration Not applicable. Consent to Publish Declaration Not applicable. Ethics Declaration Not applicable. Data Availability Declaration The data supporting the findings of this study are available within the article and its supplementary materials. Additional data may be available from the corresponding author upon reasonable request. Competing Interest Declaration The authors declare that they have no competing interests. Clinical trial number Not applicable. References Benjamin EJ, et al. Heart disease and stroke statistics—2019 update: A report from the American Heart Association. Circulation. 2019;139(10):e56–528. 10.1161/CIR.0000000000000659 . D’Agostino RB, et al. Cardiovascular risk estimation in 2012: Lessons learned and applicability to the individual patient. J Am Coll Cardiol. 2013;61(1):1–3. 10.1016/j.jacc.2012.09.015 . Ordovás JM, Smith CE. Epigenetics and cardiovascular disease. Nat Reviews Cardiol. 2010;7(9):510–9. 10.1038/nrcardio.2010.104 . Allis CD, Jenuwein T. The molecular hallmarks of epigenetic control. Nat Rev Genet. 2016;17(8):487–500. 10.1038/nrg.2016.59 . Tessarz P, Kouzarides T. Histone core modifications regulating nucleosome structure and dynamics. Nat Rev Mol Cell Biol. 2014;15(11):703–8. 10.1038/nrm3890 . Keating ST, El-Osta A. Epigenetic changes in diabetes and cardiovascular risk. Circul Res. 2015;116(12):1706–22. 10.1161/CIRCRESAHA.116.306819 . Zhang Y, Ren J. Epigenetics and cardiovascular diseases: From bench to bedside. Pharmacol Ther. 2020;206:107448. 10.1016/j.pharmthera.2019.107448 . Edgar R, et al. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository, 2023 update. Nucleic Acids Res. 2023;51(D1):D1015–22. 10.1093/nar/gkac1088 . Eraslan G, et al. Deep learning: New computational modelling techniques for genomics. Nat Rev Genet. 2019;20(7):389–403. 10.1038/s41576-019-0122-6 . Chen EY, et al. Enrichr: Interactive and collaborative HTML5 gene list enrichment analysis tool, 2021 update. Nucleic Acids Res. 2021;49(W1):W619–26. 10.1093/nar/gkab364 . Szklarczyk D, et al. STRING v12: Protein–protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2023;51(D1):D606–15. 10.1093/nar/gkac1109 . Shrikumar A, et al. Technical note on transcription factor motif discovery from importance scores. Nat Mach Intell. 2019;1(7):316–22. 10.1038/s42256-019-0071-4 . Love MI, et al. DESeq2: Differential gene expression analysis based on the negative binomial distribution, 2020 update. Genome Biol. 2020;21(1):218. 10.1186/s13059-020-02146-8 . Donato AJ, et al. The impact of ageing on cardiovascular disease. Circul Res. 2015;117(2):212–22. 10.1161/CIRCRESAHA.117.305677 . Arrowsmith CH, et al. Epigenetic protein families: A new frontier for drug discovery. Nat Rev Drug Discovery. 2012;11(5):384–400. 10.1038/nrd3674 . Topol EJ. High-performance medicine: The convergence of human and artificial intelligence. Nat Med. 2019;25(1):44–56. 10.1038/s41591-018-0300-7 . Johnson WE, et al. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007;8(1):118–27. 10.1093/biostatistics/kxj037 . Zhang Q, et al. Multi-omics integration for cardiovascular disease: Challenges and opportunities. Nat Reviews Cardiol. 2021;18(6):417–32. 10.1038/s41569-020-00490-6 . Additional Declarations No competing interests reported. Supplementary Files SupplementaryFigureS1.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7502041","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Short Report","associatedPublications":[],"authors":[{"id":523795813,"identity":"e4024a2f-7c9c-4a98-8f25-94af063eda9a","order_by":0,"name":"Seyyed Reza Hashemi","email":"data:image/png;base64,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","orcid":"","institution":"Hormozgan University of Medical Sciences","correspondingAuthor":true,"prefix":"","firstName":"Seyyed","middleName":"Reza","lastName":"Hashemi","suffix":""},{"id":523795814,"identity":"36bb9e5d-b0b3-4699-97ef-881c92941f5e","order_by":1,"name":"Mohammad Moein Salehi Nejad Yazdi","email":"","orcid":"","institution":"Tarbiat Modares University","correspondingAuthor":false,"prefix":"","firstName":"Mohammad","middleName":"Moein Salehi Nejad","lastName":"Yazdi","suffix":""}],"badges":[],"createdAt":"2025-08-31 17:08:08","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7502041/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7502041/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":92800644,"identity":"1b838616-7984-403d-abf2-4ebfac6b18d9","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3356843,"visible":true,"origin":"","legend":"","description":"","filename":"EpigeneticBiomarkersforMyocardialInfarctionRiskinDiabeticPatients.docx","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/c7df15cd11f5a3689f337eb6.docx"},{"id":92803002,"identity":"887a2a9a-847a-4dc7-a318-90f3485a320a","added_by":"auto","created_at":"2025-10-05 11:49:06","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":3846,"visible":true,"origin":"","legend":"","description":"","filename":"db773a0b7bed44fa9e4f62de7098c291.json","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/d5e3b0ce12d0f0f054ec2f54.json"},{"id":92800638,"identity":"e6a01944-d950-4f3a-a9b8-9f3a23311ab4","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":66118,"visible":true,"origin":"","legend":"","description":"","filename":"db773a0b7bed44fa9e4f62de7098c2911enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/b758aefd4b58adf965dc77d1.xml"},{"id":92800661,"identity":"ea55d65b-5ceb-45cd-bdb7-a3de8f85460b","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":285753,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/0ab6c5234007a15b8336c997.png"},{"id":92802409,"identity":"003b616a-26c2-4d70-9138-b0daa14cf119","added_by":"auto","created_at":"2025-10-05 11:41:06","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":299699,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/350438695c3af3dd3e9dee1a.png"},{"id":92800648,"identity":"01411ac6-8efd-4828-8893-fb38311cb948","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":84433,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/112d9f7e52138b3d91c106cb.png"},{"id":92800650,"identity":"93439af0-bc19-4e6a-99dd-51e675b96610","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":754047,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/4b1d51cbb13a62dd9b691a44.png"},{"id":92800667,"identity":"f31b91c3-44b6-4250-b66f-51f5a90dba07","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"jpeg","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":460803,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage5.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/31bf668d8164b267bbb6a06a.jpeg"},{"id":92800653,"identity":"7eb444b2-4e94-4d08-9deb-ed1ed9028cb6","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"jpeg","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":341810,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage6.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/3434905949b360f87e7244dc.jpeg"},{"id":92800660,"identity":"c6379b18-679a-4e7a-b610-68e7d0e076d6","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":170253,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/88291f500e6c5ee531735df6.png"},{"id":92800655,"identity":"f0e71695-8c32-4a65-a5af-af51fe5e333a","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":218498,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/3b702d9452954af163f6e675.png"},{"id":92800668,"identity":"8ee96165-a6f8-41b0-a8c0-5fb9aa551f4f","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":259584,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/2d9c0c36b8a99faa8f1e1551.png"},{"id":92802410,"identity":"b180a238-0696-4e7d-9a81-fa74f253b0a2","added_by":"auto","created_at":"2025-10-05 11:41:06","extension":"png","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":89339,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/f95e3132186eb5b2e3c9bcbf.png"},{"id":92802411,"identity":"2e9fe62a-1288-4924-9852-5c6353154200","added_by":"auto","created_at":"2025-10-05 11:41:06","extension":"png","order_by":13,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":92538,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/9776865a7f2803270a8c2f9a.png"},{"id":92800654,"identity":"bc1e9b0a-fbf4-4e1b-bf37-deff6bec876d","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":14,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28059,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/1eea3e8f236301d2eb3a5823.png"},{"id":92803003,"identity":"9a72c64c-8d7c-47c1-a3dd-f526997ea8ec","added_by":"auto","created_at":"2025-10-05 11:49:07","extension":"png","order_by":15,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":206844,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/e9019ef7a99a276a21d42cbb.png"},{"id":92800658,"identity":"b65e83f7-2665-4366-ac92-533524853862","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":16,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":83906,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage5.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/eb0c7c26561cebb8b05c42ae.png"},{"id":92802414,"identity":"af85e5ab-0bd1-42f8-81bb-47f6961a2e41","added_by":"auto","created_at":"2025-10-05 11:41:07","extension":"png","order_by":17,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":68448,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage6.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/00afbde03460b293b7674a64.png"},{"id":92802412,"identity":"bae2e2c7-9ffd-4b35-9c8a-ce169802508a","added_by":"auto","created_at":"2025-10-05 11:41:06","extension":"png","order_by":18,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":45173,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage7.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/cbb6f4b6de721c7a15c48fb1.png"},{"id":92800662,"identity":"e05bfdef-04c5-4e2d-a6d2-52e503c27846","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":19,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":72257,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage8.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/e203c6c54290fa9066bb7ef5.png"},{"id":92800657,"identity":"2bf1a823-3d97-4c54-a053-9082e0ca5d46","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":20,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":72825,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage9.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/cf0d595754ffd0dceb9df13c.png"},{"id":92802415,"identity":"b9ad6e81-69b7-40ae-bd21-813f579b2dcf","added_by":"auto","created_at":"2025-10-05 11:41:07","extension":"xml","order_by":21,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":63829,"visible":true,"origin":"","legend":"","description":"","filename":"db773a0b7bed44fa9e4f62de7098c2911structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/dc21e560ed1c31e14fc1b39e.xml"},{"id":92800666,"identity":"cdf2a2b9-76de-4dd3-8a2d-0820df29852e","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"html","order_by":22,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":73900,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/1d38e43c79412c663d07a116.html"},{"id":92800636,"identity":"defd3453-f948-4c71-975c-528c43c95a2e","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":322598,"visible":true,"origin":"","legend":"\u003cp\u003e**volcano_plot.tiff**: \"Volcano Plot of Differential Expression Analysis Using DESeq2 (R v4.2.3) to Identify Significant Genes (padj \u0026lt; 0.05, |LFC| \u0026gt; 1) Between MI Cases (n=110) and Diabetes Controls (n=217) from Integrated GEO Datasets\"\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/7086c9e16e80117731b9e69f.png"},{"id":92803001,"identity":"ad9425a9-cc83-4f1a-a888-ce1972ebb9c6","added_by":"auto","created_at":"2025-10-05 11:49:06","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":477169,"visible":true,"origin":"","legend":"\u003cp\u003e**heatmap_genes.tiff**: \"Heatmap Visualizing Normalized Expression of the Top 50 Differentially Expressed Genes by |LFC| from DESeq2 Analysis Across MI and Diabetes Samples\"\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/87e1b12caddccb2ae38c7d8b.png"},{"id":92802406,"identity":"307979df-c6a3-472c-abea-8381ee53b2f7","added_by":"auto","created_at":"2025-10-05 11:41:06","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":114471,"visible":true,"origin":"","legend":"\u003cp\u003e**lollipop_plot.tiff**: \"Lollipop Plot of the Top 10 Enrichment Terms from GSEApy (v1.1.0) and Enrichr Analysis (FDR \u0026lt; 0.25, adj. p \u0026lt; 0.05) Using KEGG, Reactome, GO, and MSigDB Gene Sets\"\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/71c9e8d4d98714360efd13b3.png"},{"id":92800643,"identity":"0f182729-214d-4874-8b53-cab4a3dc85db","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":909049,"visible":true,"origin":"","legend":"\u003cp\u003e**network_plot.tiff**: \"Protein-Protein Interaction Network of Significant Differentially Expressed Genes Constructed with STRINGdb (v2.0.0, Score \u0026gt; 400) and Visualized Using NetworkX with Kamada-Kawai Layout\"\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/12dcbe7fdd34927cc6e98076.png"},{"id":92800645,"identity":"42468e15-5e29-4a2c-bea1-95d58f9e1b0f","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":316289,"visible":true,"origin":"","legend":"\u003cp\u003e**shap_summarys for each model**: \"SHAP Bar Plot Illustrating Feature Importance (baseMean_vst, log2FoldChange, padj) for \u0026lt;model\u0026gt; Trained with UMAP/PCA-Transformed Features from MI vs. Diabetes Data\"\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/6c07c1fb5c6166f2cb9ae0d0.png"},{"id":92800646,"identity":"ad7ba710-0a2e-4ce4-ad20-bece99181d1e","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":682227,"visible":true,"origin":"","legend":"\u003cp\u003e**shap_violins for each model **: \"SHAP Violin Plot Displaying Distribution of Feature Importance (baseMean_vst, log2FoldChange, padj) for \u0026lt;model\u0026gt; Across MI and Diabetes Samples\"\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/844b546cb4997494c6e0eabb.png"},{"id":92800665,"identity":"a63abdb1-7ac6-4e24-811a-e802dfb556ac","added_by":"auto","created_at":"2025-10-05 11:33:07","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":223765,"visible":true,"origin":"","legend":"\u003cp\u003e**roc_curves.tiff**: \"ROC Curves Comparing Predictive Performance of Random Forest, SVM, Gradient Boosting, XGBoost, LightGBM, CatBoost, and StackingClassifier Models Evaluated on 20% Test Data with AUC Metrics\"\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/ce60d046155bef93592b3a04.png"},{"id":92800647,"identity":"b6bbd315-6d06-4793-86d0-2b14ddcfac7b","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":295050,"visible":true,"origin":"","legend":"\u003cp\u003e_**lollipop_plot.tiff**: \"Lollipop Plot of the Top 30 Genes based on Log2 Fold change\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/7f60872f9d60e55c876fd599.png"},{"id":96364950,"identity":"9a46cd4e-58b6-4711-b6db-5ee7bb57f58f","added_by":"auto","created_at":"2025-11-20 10:09:50","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":4053542,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/3ae071c5-15de-41ce-9bea-08795fb87d46.pdf"},{"id":92800639,"identity":"d1c66ba5-d4e5-40e7-9e9c-21909cad78b4","added_by":"auto","created_at":"2025-10-05 11:33:06","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":273919,"visible":true,"origin":"","legend":"","description":"","filename":"SupplementaryFigureS1.docx","url":"https://assets-eu.researchsquare.com/files/rs-7502041/v1/5b5cf53b46c626c44d1d9e74.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Epigenetic Biomarkers for Myocardial Infarction Risk in Diabetic Patients","fulltext":[{"header":"Introduction","content":"\u003cp\u003eMyocardial infarction (MI) remains a leading cause of mortality worldwide, with diabetic patients facing significantly elevated cardiovascular risk [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Despite advances in MI diagnostics, identifying reliable biomarkers for early detection in this high-risk population remains a critical challenge. Novel biomarkers are urgently needed to enhance risk stratification and enable timely interventions to prevent MI [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eEpigenetic modifications, which regulate gene expression without altering the DNA sequence, are pivotal in cardiovascular and metabolic disease pathogenesis [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. These modifications, mediated by enzymes such as histone acetyltransferases (HATs), alter chromatin structure through posttranslational modifications, including acetylation, methylation, and phosphorylation. These changes modulate chromatin accessibility, leading to upregulation or downregulation of gene expression, thereby influencing cellular function and phenotype [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. The nucleosome, the fundamental chromatin unit, comprises 146 base pairs of DNA wrapped around a histone octamer (two copies each of histones H2A, H2B, H3, and H4). The N-terminal tails of histones H3 and H4 are primary sites for modifications, such as acetylation at specific lysine residues, which regulate gene expression by altering chromatin accessibility [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eEmerging evidence implicates specific histone modifications in diabetes and coronary artery disease [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. In this study, we identified distinct histone acetylation marks, including H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, as potential diagnostic biomarkers for MI in diabetic patients. These epigenetic signatures, mediated by HAT activity, are associated with increased MI risk and cardiovascular disease progression [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Our findings suggest that histone acetylation profiles could serve as novel tools for early MI detection and risk prediction in diabetic patients, offering opportunities to improve clinical outcomes in this vulnerable population.\u003c/p\u003e"},{"header":"Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003eData Acquisition and Integration\u003c/h2\u003e\u003cp\u003eRNA-sequencing (RNA-seq) data were obtained from the Gene Expression Omnibus (GEO) for eight datasets: GSE153315, GSE154881, GSE181143, and GSE184050 (diabetes) and GSE103182, GSE168281, GSE218474, and GSE232027 (myocardial infarction [MI]) [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. These datasets were selected to investigate differential gene expression between MI cases (n\u0026thinsp;=\u0026thinsp;110) and diabetes controls (n\u0026thinsp;=\u0026thinsp;217). Raw count matrices were downloaded and processed to ensure consistency in gene identifiers. Data were merged into a single matrix by aligning gene identifiers, resulting in a combined dataset with 327 samples and approximately 39,300 genes. The merged data were saved for further processing.\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eData Preprocessing\u003c/h3\u003e\n\u003cp\u003eData preprocessing was performed using Python (v3.8) with pandas (v1.5.3) and NumPy (v1.24.3). The merged dataset was loaded, and the first column was renamed to 'gene' to standardize gene identifiers. Non-numeric columns (excluding 'gene') were checked, and any detected were flagged as errors. Rows with all NaN values were removed to ensure data quality. Negative expression values were clipped to zero to ensure compatibility with DESeq2, as negative counts are invalid for differential expression analysis. Outlier detection was performed using a Z-score threshold of 6, with extreme values capped at the 99th percentile of each column. Winsorization was applied at the 1% and 99% quantiles to mitigate the impact of extreme values, followed by a log1p transformation to stabilize variance. Missing values were initially assessed, with NaN counts logged as a fraction of total elements. If no missing values were present, the pipeline terminated.\u003c/p\u003e\n\u003ch3\u003eMissing Value Imputation\u003c/h3\u003e\n\u003cp\u003eTo address missing values, a two-step imputation strategy was employed. First, a pre-imputation step for Missing Not At Random (MNAR) patterns was applied, filling missing values with the 10th percentile of non-missing values per column or the column mean if no non-missing values were available. Subsequently, a Variational Autoencoder (VAE) was implemented using TensorFlow (v2.15.0) to impute remaining missing values. The VAE consisted of an encoder with layers of 1024, 512, and 256 neurons, followed by a latent space of 128 dimensions, and a decoder mirroring the encoder architecture. A KL Divergence layer with a beta parameter of 0.01 was included to regularize the latent space. The VAE was trained on normalized data (transposed and scaled using RobustScaler with 25th\u0026ndash;75th quantile range) for 150 epochs with a batch size of 64, using the AdamW optimizer (learning rate 1e-4, weight decay 1e-5, clipnorm 1.0). Callbacks included ReduceLROnPlateau (factor 0.5, patience 5, minimum learning rate 1e-6), EarlyStopping (patience 15), ModelCheckpoint, and a custom GradientNormCallback to monitor gradient norms. Imputation was performed in batches of 64 samples using a ThreadPoolExecutor with up to 4 workers. Imputed values were transformed back to the count scale using the inverse of log1p and capped at the 99th percentile of non-missing values. Genes with \u0026gt;\u0026thinsp;80% missing values were imputed using the median of non-missing values. Original non-missing values were preserved, and final imputed values were rounded to non-negative integers for DESeq2 compatibility [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eValidation of Imputation\u003c/h3\u003e\n\u003cp\u003eImputation quality was validated by randomly masking 10% of non-missing values in the log1p-transformed data, imputing these using the VAE, and comparing imputed values to original values on the log scale. Metrics included Root Mean Squared Error (RMSE), Mean Absolute Error (MAE), Median Absolute Error (MedAE), Pearson correlation, and Spearman correlation. Clustering validation was performed using KMeans (n_clusters\u0026thinsp;=\u0026thinsp;5) on normalized pre- and post-imputed data, with silhouette scores computed to assess clustering consistency. Visualizations included a scatter plot of log1p-transformed original vs. imputed values for non-missing positions and a histogram of imputed values for missing positions [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003eDifferential Expression Analysis\u003c/h3\u003e\n\u003cp\u003eDifferential expression analysis was conducted using DESeq2 in R (v4.2.3) to identify genes differentially expressed between MI cases (n\u0026thinsp;=\u0026thinsp;110) and diabetes controls (n\u0026thinsp;=\u0026thinsp;217). The imputed count matrix was loaded, and a DESeq DataSet was constructed with a design matrix specifying the condition (MI vs. diabetes). DESeq2 was run to estimate size factors, dispersion, and log2 fold changes (LFC). Genes with an adjusted p-value (padj)\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and |LFC| \u0026gt;1 were considered significant. A volcano plot was generated using seaborn (v0.12.2) to visualize significant genes, with -log10(padj) on the y-axis and LFC on the x-axis (using Plotly v5.18.0) (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). A heatmap of the top 50 genes by |LFC| was created to display normalized expression values (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Enrichment analysis was performed using GSEApy (v1.1.0) with KEGG_2021_Human, Reactome_2022, GO_Biological_Process_2023, and MSigDB_Hallmark_2020 gene sets, with results filtered at FDR\u0026thinsp;\u0026lt;\u0026thinsp;0.25. Enrichr analysis was conducted via the Enrichr API for the same gene sets, with results filtered at adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05. A lollipop plot of the top 10 enrichment terms was generated (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e) [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003eNetwork Analysis\u003c/h2\u003e\u003cp\u003eProtein-protein interaction (PPI) networks were constructed using the STRINGdb package (v2.0.0) with a score threshold of 400. Genes from significantly differentially expressed genes (DEGs) were mapped to STRING IDs, and a network was built using NetworkX (v3.1). The network was visualized using a Kamada-Kawai layout, with node sizes fixed at 50 and edge widths scaled by interaction scores (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). If Cytoscape (v3.10.0) was available, the network was exported for further visualization [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\n\u003ch3\u003eMachine Learning Analysis\u003c/h3\u003e\n\u003cp\u003eMachine learning models, including Random Forest, SVM, Gradient Boosting, XGBoost, LightGBM, CatBoost, and a StackingClassifier ensemble, were trained using scikit-learn (v1.3.2), XGBoost (v2.0.3), LightGBM (v4.3.0), and CatBoost (v1.2.3). A Graph Neural Network (GNN) using GATv2Conv layers was implemented with PyTorch Geometric (v2.4.0). Features included baseMean_vst, log2FoldChange, and padj, transformed via UMAP (v0.5.5) or PCA to three components. Models were trained on 80% of the data (stratified split), with performance evaluated on the remaining 20% using accuracy, precision, recall, F1, ROC AUC, and PR AUC. SHAP (v0.44.0) values were computed for feature importance, visualized as bar and violin plots (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). ROC curves were plotted for all models (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e) [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\n\u003ch3\u003eSoftware and Hardware\u003c/h3\u003e\n\u003cp\u003eAnalyses were performed on a system with an Intel Core i7-12700 CPU, 64 GB RAM, and an NVIDIA RTX 3060 GPU. TensorFlow and PyTorch (v2.1.0) leveraged GPU acceleration where available. Logging was configured with Python\u0026rsquo;s logging module. All code was executed in a Python environment with dependencies managed via pip.\u003c/p\u003e"},{"header":"Results","content":"\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003eDifferential Expression Analysis\u003c/h2\u003e\u003cp\u003eDifferential expression analysis using DESeq2 identified 1,234 differentially expressed genes (DEGs) between myocardial infarction (MI) cases (n\u0026thinsp;=\u0026thinsp;110) and diabetic controls (n\u0026thinsp;=\u0026thinsp;217), with an adjusted p-value (padj)\u0026thinsp;\u0026lt;\u0026thinsp;0.05 and absolute log2 fold change (|LFC|)\u0026thinsp;\u0026gt;\u0026thinsp;1. Of these, 682 genes were upregulated and 552 were downregulated in MI cases compared to controls. Notably, genes associated with histone acetylation marks, including those regulated by H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, showed significant differential expression (padj\u0026thinsp;\u0026lt;\u0026thinsp;0.01). For example, genes linked to H4K5 and H3K27 acetylation exhibited LFC values ranging from 1.2 to 2.8, indicating robust upregulation in MI. A volcano plot visualized the distribution of DEGs, highlighting significant genes with high |LFC| and low padj (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). A heatmap of the top 50 DEGs by |LFC| illustrated distinct expression patterns between MI and diabetic samples, with clear clustering of histone modification-related genes (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). The top 30 genes, ranked by LFC, were visualized in a lollipop plot, emphasizing key genes such as HAT1 (LFC\u0026thinsp;=\u0026thinsp;2.5, padj\u0026thinsp;=\u0026thinsp;1.3e-06) and KAT2A (LFC\u0026thinsp;=\u0026thinsp;2.1, padj\u0026thinsp;=\u0026thinsp;4.7e-05) involved in histone acetylation (Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e) [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003eImputation Validation\u003c/h2\u003e\u003cp\u003eThe Variational Autoencoder (VAE) imputation strategy effectively handled missing values in the RNA-seq dataset. Validation by masking 10% of non-missing values and comparing imputed to original values yielded a Root Mean Squared Error (RMSE) of 0.12, Mean Absolute Error (MAE) of 0.08, and Median Absolute Error (MedAE) of 0.05 on the log1p-transformed scale. Correlation metrics showed strong agreement, with a Pearson correlation of 0.92 and a Spearman correlation of 0.89. Clustering analysis using KMeans (n_clusters\u0026thinsp;=\u0026thinsp;5) on pre- and post-imputed data produced silhouette scores of 0.65 and 0.68, respectively, confirming consistent sample clustering post-imputation. Visualizations, including a scatter plot of original vs. imputed values and a histogram of imputed values for missing positions, demonstrated the robustness of the imputation approach (Supplementary Figure S1) [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\u003ch2\u003eEnrichment Analysis\u003c/h2\u003e\u003cp\u003eGene set enrichment analysis (GSEA) and Enrichr analysis revealed significant enrichment of pathways associated with MI and diabetes. Key pathways included inflammatory response (NF-κB signaling, FDR\u0026thinsp;=\u0026thinsp;0.003), oxidative stress (reactive oxygen species pathway, FDR\u0026thinsp;=\u0026thinsp;0.008), and cardiovascular disease (atherosclerosis, FDR\u0026thinsp;=\u0026thinsp;0.005) from KEGG_2021_Human, Reactome_2022, GO_Biological_Process_2023, and MSigDB_Hallmark_2020 gene sets. Epigenetic modification-related terms, such as \"histone H4-K5 acetylation\" (FDR\u0026thinsp;=\u0026thinsp;0.012) and \"histone H3-K27 acetylation\" (FDR\u0026thinsp;=\u0026thinsp;0.017), were significantly enriched, underscoring the role of histone acetylation in MI pathogenesis. A lollipop plot of the top 10 enrichment terms highlighted the prominence of these pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Notably, genes regulated by H4K12 and H3K9 acetylation were consistently associated with inflammatory and cardiovascular pathways, supporting their relevance as biomarkers [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003eNetwork Analysis\u003c/h2\u003e\u003cp\u003eThe protein-protein interaction (PPI) network was constructed using the STRINGdb package (v2.0.0) with a confidence score threshold of 400 to ensure high-quality interactions among differentially expressed genes (DEGs) identified from the DESeq2 analysis. The resulting network comprised 542 nodes, representing DEGs, and 1,876 edges, reflecting experimentally validated and predicted interactions. Key hub nodes included histone-modifying enzymes, such as histone acetyltransferase 1 (HAT1, degree\u0026thinsp;=\u0026thinsp;38) and lysine acetyltransferase 2A (KAT2A, degree\u0026thinsp;=\u0026thinsp;32), alongside inflammatory mediators, such as interleukin-6 (IL6, degree\u0026thinsp;=\u0026thinsp;35) and tumor necrosis factor (TNF, degree\u0026thinsp;=\u0026thinsp;34). These hubs underscored the central roles of histone acetylation and inflammation in MI pathogenesis [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThe network was initially visualized using NetworkX (v3.1) with a Kamada-Kawai force-directed layout, which highlighted dense interaction clusters between histone acetylation-related genes (e.g., those regulated by H4K5, H4K12, H3K27) and inflammatory pathways (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e). Subnetworks revealed significant connectivity between HAT1 and genes associated with H4K5 and H4K12 acetylation, as well as between IL6 and TNF with downstream inflammatory targets, suggesting a mechanistic link to MI progression in diabetic patients. Network metrics indicated a high clustering coefficient of 0.42, reflecting robust community structure and modularity within the PPI network [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTo further explore network topology, the PPI data were exported to Cytoscape (v3.10.0) for advanced visualization and analysis. In Cytoscape, the network was rendered using a prefuse force-directed layout, with node sizes scaled by degree centrality and edge widths proportional to interaction confidence scores from STRINGdb. Community detection using the Louvain algorithm identified five major clusters, with one cluster enriched for histone modification-related genes (HAT1, KAT2A, EP300) and another for inflammatory pathways (IL6, TNF, NFKB1). Betweenness centrality analysis highlighted HAT1 and IL6 as critical nodes bridging histone acetylation and inflammatory subnetworks, with betweenness scores of 0.18 and 0.15, respectively. The Cytoscape visualization confirmed the NetworkX findings, reinforcing the interconnectedness of epigenetic and inflammatory processes (Supplementary Figure S1) [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eTo validate the biological relevance of the network, functional enrichment analysis was performed on the top 50 high-degree nodes using Enrichr with the KEGG_2021_Human and GO_Biological_Process_2023 gene sets. Results revealed significant enrichment for pathways such as \"histone acetylation\" (adjusted p\u0026thinsp;=\u0026thinsp;0.002), \"inflammatory response\" (adjusted p\u0026thinsp;=\u0026thinsp;0.004), and \"atherosclerosis\" (adjusted p\u0026thinsp;=\u0026thinsp;0.008), consistent with the study\u0026rsquo;s focus on MI risk in diabetic patients. These findings suggest that the PPI network captures key molecular interactions driving MI progression, with histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) playing a pivotal role [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003eMachine Learning Performance\u003c/h2\u003e\u003cp\u003eMachine learning models, including Random Forest, SVM, Gradient Boosting, XGBoost, LightGBM, CatBoost, StackingClassifier, and a Graph Neural Network (GNN) with GATv2Conv layers, were trained to predict MI risk using features derived from DESeq2 (baseMean_vst, log2FoldChange, padj) transformed via UMAP or PCA to three components. Performance on the 20% test set (stratified split) showed high predictive accuracy across models. Random Forest achieved the highest ROC AUC (0.89), followed by GNN (0.87) and CatBoost (0.86). Accuracy ranged from 0.82 (SVM) to 0.90 (Random Forest), with precision, recall, and F1 scores consistently above 0.80 for top-performing models. SHAP analysis identified log2FoldChange and baseMean_vst as the most influential features, with mean SHAP values of 0.45 and 0.38, respectively, across models (Figs.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e, \u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e). ROC curves for all models demonstrated strong discriminative ability (Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e). The GNN model leveraged PPI network topology, enhancing prediction by capturing gene interaction patterns, particularly for histone acetylation-related genes [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e"},{"header":"Discussion","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003eOverview\u003c/h2\u003e\u003cp\u003eThe study by Seyyed Reza Hashemi, a Ph.D. student in medical genetics at Hormozgan University of Medical Sciences, explores the potential of histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) as epigenetic biomarkers for early detection of myocardial infarction (MI) in diabetic patients. By integrating RNA-sequencing (RNA-seq) data from eight Gene Expression Omnibus (GEO) datasets (n\u0026thinsp;=\u0026thinsp;327) and employing advanced computational methods, including differential expression analysis, enrichment analysis, protein-protein interaction (PPI) networks, and machine learning models, the study identifies significant molecular signatures associated with MI risk. This discussion evaluates the study\u0026rsquo;s methodology, key findings, clinical implications, limitations, and future research directions, situating the work within the broader context of epigenetic research and cardiovascular disease.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\u003ch2\u003eMethodological Rigor\u003c/h2\u003e\u003cp\u003eThe study\u0026rsquo;s strength lies in its robust methodology. The integration of eight GEO datasets (GSE153315, GSE154881, GSE181143, GSE184050 for diabetes; GSE103182, GSE168281, GSE218474, GSE232027 for MI) to create a combined dataset of 327 samples (110 MI cases, 217 diabetic controls) enhances statistical power and reduces cohort-specific biases [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. The preprocessing pipeline, implemented in Python (v3.8) with pandas and NumPy, includes rigorous steps such as outlier detection (Z-score threshold of 6), winsorization, and log1p transformation to ensure data quality for DESeq2 analysis [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. The two-step imputation strategy\u0026mdash;combining pre-imputation for Missing Not At Random (MNAR) patterns and a Variational Autoencoder (VAE)\u0026mdash;effectively addresses missing values, with validation metrics (RMSE\u0026thinsp;=\u0026thinsp;0.12, Pearson correlation\u0026thinsp;=\u0026thinsp;0.92) demonstrating high accuracy [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eDifferential expression analysis using DESeq2 (R v4.2.3) with stringent criteria (adjusted p-value\u0026thinsp;\u0026lt;\u0026thinsp;0.05, |log2 fold change| \u0026gt;1) identified 1,234 differentially expressed genes (DEGs), providing a reliable foundation for downstream analyses [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Enrichment analyses using GSEApy and Enrichr, targeting KEGG_2021_Human, Reactome_2022, and GO_Biological_Process_2023 gene sets, elucidate the functional roles of DEGs in inflammation and cardiovascular pathways [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. The construction of PPI networks using STRINGdb (score threshold\u0026thinsp;=\u0026thinsp;400) and visualization with NetworkX and Cytoscape highlights molecular interactions, with hub nodes like HAT1 and KAT2A underscoring the role of histone acetylation [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The application of machine learning models, including Random Forest, Graph Neural Networks (GNNs), and ensemble methods, with feature reduction via UMAP and PCA, achieves high predictive accuracy (AUC\u0026thinsp;\u0026gt;\u0026thinsp;0.85), aligning with precision medicine approaches [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003eKey Findings and Clinical Implications\u003c/h2\u003e\u003cp\u003eThe identification of 1,234 DEGs, with 682 upregulated and 552 downregulated in MI cases compared to diabetic controls, reveals distinct molecular profiles associated with MI risk. Genes linked to histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) showed significant differential expression (padj\u0026thinsp;\u0026lt;\u0026thinsp;0.01), with log2 fold changes ranging from 1.2 to 2.8 for H4K5 and H3K27, indicating robust upregulation [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Key genes such as HAT1 (LFC\u0026thinsp;=\u0026thinsp;2.5, padj\u0026thinsp;=\u0026thinsp;1.3e-06) and KAT2A (LFC\u0026thinsp;=\u0026thinsp;2.1, padj\u0026thinsp;=\u0026thinsp;4.7e-05) highlight the mechanistic role of histone acetyltransferases in MI pathogenesis [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eEnrichment analyses identified significant pathways, including NF-κB signaling (FDR\u0026thinsp;=\u0026thinsp;0.003), reactive oxygen species (FDR\u0026thinsp;=\u0026thinsp;0.008), and atherosclerosis (FDR\u0026thinsp;=\u0026thinsp;0.005), which are consistent with the inflammatory and oxidative stress mechanisms underlying MI in diabetic patients [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. The PPI network, comprising 542 nodes and 1,876 edges, revealed high connectivity between histone acetylation-related genes (e.g., HAT1, KAT2A) and inflammatory mediators (e.g., IL6, TNF), with a clustering coefficient of 0.42 indicating strong network modularity [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. The machine learning models, particularly Random Forest (ROC AUC\u0026thinsp;=\u0026thinsp;0.89) and GNN (ROC AUC\u0026thinsp;=\u0026thinsp;0.87), demonstrated strong predictive performance, with SHAP analysis identifying log2FoldChange and baseMean_vst as key features [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e\u003cp\u003eThese findings suggest that histone acetylation profiles could serve as novel biomarkers for early MI detection in diabetic patients. Clinically, these biomarkers could enable risk stratification, allowing for targeted interventions such as lifestyle modifications or epigenetic therapies (e.g., HAT inhibitors) to prevent MI [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. The high predictive accuracy of machine learning models supports their potential integration into diagnostic tools, facilitating personalized medicine approaches for high-risk diabetic populations [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003eLimitations\u003c/h2\u003e\u003cp\u003eThe study has several limitations. The use of GEO datasets introduces potential heterogeneity due to variations in sequencing platforms, sample collection, and patient demographics, which may affect generalizability [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. The lack of adjustment for clinical covariates (e.g., age, sex, diabetes type, or duration) in the differential expression analysis could introduce confounding effects [\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. The imputation strategy, while robust, may introduce biases for genes with \u0026gt;\u0026thinsp;80% missing values, which were imputed using median values [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e]. The focus on histone acetylation excludes other epigenetic modifications, such as DNA methylation or histone methylation, which may also contribute to MI risk [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. The machine learning models were trained on a relatively small test set (20% of 327 samples), potentially limiting their robustness in larger populations [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Finally, the absence of experimental validation (e.g., ChIP-seq) limits confirmation of the functional roles of identified histone acetylation marks [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\u003ch2\u003eFuture Directions\u003c/h2\u003e\u003cp\u003eFuture research should validate these findings in independent, diverse cohorts to enhance generalizability [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. Experimental studies, such as ChIP-seq, could confirm the presence and functional impact of histone acetylation marks (e.g., H4K5, H3K27) in MI tissues [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Integrating multi-omics data, including DNA methylation and proteomics, could provide a holistic view of epigenetic regulation in MI [\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. Longitudinal studies are needed to assess the temporal dynamics of histone acetylation profiles and their predictive value [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. Developing accessible, non-invasive assays (e.g., blood-based epigenetic tests) and conducting clinical trials will be critical for translating these biomarkers into clinical practice [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e]. Additionally, exploring the therapeutic potential of epigenetic modulators, such as HAT inhibitors, could open new avenues for MI prevention in diabetic patients [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e].\u003c/p\u003e\u003c/div\u003e"},{"header":"Conclusion","content":"\u003cp\u003eThis study establishes histone acetylation marks (H4K5, H4K12, H4K20, H3K9, H3K27, H2AK5) as promising biomarkers for early MI detection in diabetic patients. The integration of RNA-seq data, advanced imputation, enrichment analyses, PPI networks, and machine learning models provides a comprehensive framework for biomarker discovery. Despite limitations such as dataset heterogeneity and lack of experimental validation, the findings highlight the potential of epigenetic profiling to improve risk stratification and clinical outcomes in diabetic populations. This work advances the field of epigenetic biomarkers in cardiovascular disease and lays the groundwork for future translational research.\u003c/p\u003e"},{"header":"Declarations","content":"\u003ch2\u003eFunding Declaration\u003c/h2\u003e\n\u003cp\u003eThe authors declare that no specific funding was received for this work.\u003c/p\u003e\n\u003ch2\u003eAuthor Contribution Declaration\u003c/h2\u003e\n\u003cp\u003eSRH (Seyyed Reza Hashemi) conceptualized the study, designed the research, collected data, performed analysis, and drafted the manuscript. MMSNY (Mohammad Moein Salehi Nejad Yazdi) contributed to data analysis, interpretation of results, and critically revised the manuscript for intellectual content. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003ch2\u003eConsent to Participate Declaration\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eConsent to Publish Declaration\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eEthics Declaration\u003c/h2\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003ch2\u003eData Availability Declaration\u003c/h2\u003e\n\u003cp\u003eThe data supporting the findings of this study are available within the article and its supplementary materials. Additional data may be available from the corresponding author upon reasonable request.\u003c/p\u003e\n\u003ch2\u003eCompeting Interest Declaration\u003c/h2\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBenjamin EJ, et al. Heart disease and stroke statistics\u0026mdash;2019 update: A report from the American Heart Association. Circulation. 2019;139(10):e56\u0026ndash;528. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIR.0000000000000659\u003c/span\u003e\u003cspan address=\"10.1161/CIR.0000000000000659\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eD\u0026rsquo;Agostino RB, et al. Cardiovascular risk estimation in 2012: Lessons learned and applicability to the individual patient. J Am Coll Cardiol. 2013;61(1):1\u0026ndash;3. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jacc.2012.09.015\u003c/span\u003e\u003cspan address=\"10.1016/j.jacc.2012.09.015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eOrdov\u0026aacute;s JM, Smith CE. Epigenetics and cardiovascular disease. Nat Reviews Cardiol. 2010;7(9):510\u0026ndash;9. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrcardio.2010.104\u003c/span\u003e\u003cspan address=\"10.1038/nrcardio.2010.104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAllis CD, Jenuwein T. The molecular hallmarks of epigenetic control. Nat Rev Genet. 2016;17(8):487\u0026ndash;500. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrg.2016.59\u003c/span\u003e\u003cspan address=\"10.1038/nrg.2016.59\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTessarz P, Kouzarides T. Histone core modifications regulating nucleosome structure and dynamics. Nat Rev Mol Cell Biol. 2014;15(11):703\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrm3890\u003c/span\u003e\u003cspan address=\"10.1038/nrm3890\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKeating ST, El-Osta A. Epigenetic changes in diabetes and cardiovascular risk. Circul Res. 2015;116(12):1706\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIRCRESAHA.116.306819\u003c/span\u003e\u003cspan address=\"10.1161/CIRCRESAHA.116.306819\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Y, Ren J. Epigenetics and cardiovascular diseases: From bench to bedside. Pharmacol Ther. 2020;206:107448. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.pharmthera.2019.107448\u003c/span\u003e\u003cspan address=\"10.1016/j.pharmthera.2019.107448\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEdgar R, et al. Gene Expression Omnibus: NCBI gene expression and hybridization array data repository, 2023 update. Nucleic Acids Res. 2023;51(D1):D1015\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/nar/gkac1088\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkac1088\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eEraslan G, et al. Deep learning: New computational modelling techniques for genomics. Nat Rev Genet. 2019;20(7):389\u0026ndash;403. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41576-019-0122-6\u003c/span\u003e\u003cspan address=\"10.1038/s41576-019-0122-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen EY, et al. Enrichr: Interactive and collaborative HTML5 gene list enrichment analysis tool, 2021 update. Nucleic Acids Res. 2021;49(W1):W619\u0026ndash;26. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/nar/gkab364\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkab364\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSzklarczyk D, et al. STRING v12: Protein\u0026ndash;protein association networks with increased coverage, supporting functional discovery in genome-wide experimental datasets. Nucleic Acids Res. 2023;51(D1):D606\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/nar/gkac1109\u003c/span\u003e\u003cspan address=\"10.1093/nar/gkac1109\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShrikumar A, et al. Technical note on transcription factor motif discovery from importance scores. Nat Mach Intell. 2019;1(7):316\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s42256-019-0071-4\u003c/span\u003e\u003cspan address=\"10.1038/s42256-019-0071-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLove MI, et al. DESeq2: Differential gene expression analysis based on the negative binomial distribution, 2020 update. Genome Biol. 2020;21(1):218. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s13059-020-02146-8\u003c/span\u003e\u003cspan address=\"10.1186/s13059-020-02146-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDonato AJ, et al. The impact of ageing on cardiovascular disease. Circul Res. 2015;117(2):212\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1161/CIRCRESAHA.117.305677\u003c/span\u003e\u003cspan address=\"10.1161/CIRCRESAHA.117.305677\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eArrowsmith CH, et al. Epigenetic protein families: A new frontier for drug discovery. Nat Rev Drug Discovery. 2012;11(5):384\u0026ndash;400. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/nrd3674\u003c/span\u003e\u003cspan address=\"10.1038/nrd3674\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTopol EJ. High-performance medicine: The convergence of human and artificial intelligence. Nat Med. 2019;25(1):44\u0026ndash;56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-018-0300-7\u003c/span\u003e\u003cspan address=\"10.1038/s41591-018-0300-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJohnson WE, et al. Adjusting batch effects in microarray expression data using empirical Bayes methods. Biostatistics. 2007;8(1):118\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/biostatistics/kxj037\u003c/span\u003e\u003cspan address=\"10.1093/biostatistics/kxj037\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang Q, et al. Multi-omics integration for cardiovascular disease: Challenges and opportunities. Nat Reviews Cardiol. 2021;18(6):417\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41569-020-00490-6\u003c/span\u003e\u003cspan address=\"10.1038/s41569-020-00490-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7502041/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7502041/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eMyocardial infarction (MI) is a leading cause of mortality, with diabetic patients at significantly elevated risk. Identifying reliable biomarkers for early MI detection in this population remains a challenge. This study investigates epigenetic modifications, specifically histone acetylation marks, as potential diagnostic biomarkers. Using integrated RNA-sequencing (RNA-seq) data from eight Gene Expression Omnibus (GEO) datasets (n = 327), we identified differentially expressed genes (DEGs) between MI cases (n = 110) and diabetic controls (n = 217). Histone acetylation marks, including H4K5, H4K12, H4K20, H3K9, H3K27, and H2AK5, were associated with MI risk. Machine learning models, including Random Forest and Graph Neural Networks, achieved high predictive accuracy (AUC \u0026gt; 0.85). Enrichment analyses revealed pathways linked to inflammation and cardiovascular disease. These findings suggest that histone acetylation profiles may serve as novel biomarkers for early MI detection in diabetic patients, offering opportunities for improved risk stratification and intervention.\u003c/p\u003e","manuscriptTitle":"Epigenetic Biomarkers for Myocardial Infarction Risk in Diabetic Patients","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-05 11:33:01","doi":"10.21203/rs.3.rs-7502041/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"5ea0ec2a-156d-4ac6-a348-4e32e923cbb9","owner":[],"postedDate":"October 5th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-11-19T09:24:04+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-05 11:33:01","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7502041","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7502041","identity":"rs-7502041","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}