Abstract

Background

   Ferroptosis is a cell death process that depends on iron and reactive
   oxygen species. It significantly contributes to cardiovascular
   diseases. However, its exact role in ischemic cardiomyopathy (ICM) is
   still unclear.

Methods

   Using bioinformatics methods, we identified new molecular targets
   associated with ferroptosis in ICM and conducted various
   analyses—including correlation analysis, pathway enrichment analysis,
   protein interaction network construction, and analysis of transcription
   factor and drug interactions, to reveal the potential mechanisms behind
   these genes.

Results

   We evaluated two independent training sets of ICM, [37]GSE57338 and
   [38]GSE5406, comprising 203 ICM samples, and validation sets
   [39]GSE76701 to examine differentially expressed genes (DEGs) related
   to ferroptosis. After extracting the intersection of the gene sets and
   ferroptosis-related genes, 53 DEGs were identified. Enrichment analyses
   showed that the alterations in ferroptosis-related DEGs were mainly
   enriched in oxidative stress response, and immune-related pathways.
   Furthermore, 11 hub genes were identified using protein–protein
   interaction network analysis. The key interactions between 11 hub genes
   were more pronounced in protein localization during ICM development. In
   addition, we construct a hub gene and transcription factor interaction
   network and a small molecule drug-gene interaction network. We found
   that among these hub genes, the N-acetylneuraminate outer membrane
   channel(NANC) gene is positively correlated with most of the
   small-molecule drugs used to treat ICM, and its high expression might
   increase resistance.

Conclusions

   Ferroptosis exists in ICM and and is associated with oxidative stress.
   This association suggests that ferroptosis may facilitate the
   progression of ICM.

   Keywords: Ferroptosis, Oxidative stress, Ischemic cardiomyopathy,
   Bioinformatics, Differentially expressed genes

1. Introduction

   Ischemic cardiomyopathy (ICM) is a serious type of coronary heart
   disease caused by widespread coronary atherosclerosis, which leads to
   ongoing myocardial ischemia and significant myocardial fibrosis. In the
   end, 11 % to 45 % of patients will experience severe heart failure,
   which greatly affects their overall clinical prognosis [40][1]. The
   pathophysiology of ICM is marked by its complex and heterogeneous
   characteristics, with coronary artery stenosis identified as the
   primary underlying mechanism.This stenosis causes ischemia and hypoxia
   in cardiomyocytes, which could lead to cellular death, fibrosis, and
   scar formation in the myocardial tissue. As a result, both the systolic
   and diastolic functions of the heart are severely impaired [41][2].The
   preservation of coronary artery integrity is crucial for the proper
   physiological functioning of cardiomyocytes. Several factors contribute
   to the development of coronary artery stenosis, with atherosclerosis
   being a major contributor. In cases of coronary artery stenosis, the
   blood supply to cardiomyocytes is significantly reduced, leading to
   metabolic disturbances such as lactic acid accumulation, which further
   worsens cellular injury. This compromised state could negatively impact
   cellular membrane integrity and the functionality of ion channels,
   potentially resulting in serious complications, including arrhythmias
   and even sudden cardiac death. Prolonged ischemic oxygen deprivation
   could result in apoptosis or necrosis of myocardial cells, which
   severely affects cardiac function and might lead to heart failure
   [42][3]. ICM presents a critical health challenge, with its global
   prevalence increasing, alongside a troubling trend of earlier onset
   among affected individuals [43][4]. Additionally, ICM has one of the
   highest mortality rates among chronic diseases, largely due to the lack
   of effective treatment options [44][5].

   Current treatment options for ICM include medication, percutaneous
   coronary intervention (PCI), and coronary artery bypass grafting
   (CABG). Studies show that both PCI and CABG significantly improve heart
   function. This is especially true for individuals with viable heart
   tissue [45][4]. Advances in technology have led to new treatment
   options, including mesenchymal stromal/stem cell (MSC)-based therapies
   [46][6]. While drug therapy is fundamental, it has side effects and
   cannot reverse myocardial injury. In contrast, PCI can restore blood
   flow to the heart, but it is limited to the coronary arteries and
   carries a risk of restenosis. Alternatively, CABG provides an
   alternative pathway for blood flow through surgery, though it involves
   significant risks and lengthy recovery times. Meanwhile, emerging
   therapies based on MSCs show promise, but they are still in the
   research phase and face challenges such as immune rejection and high
   costs. Despite these advancements, the prognosis for patients with ICM
   is poor, with a five-year survival rate estimated at approximately
   50–60 % [47][7]. Several key factors contribute to poor outcomes in ICM
   patients, including diabetes mellitus, left ventricular ejection
   fraction (LVEF), and a history of heart failure-related
   hospitalizations [48][8]. Hence, early diagnosis and effective
   management strategies are imperative to mitigate morbidity and
   mortality associated with this condition.

   ICM develops through various mechanisms, such as myocardial infarction
   due to ischemia, compensatory hypertrophy in unaffected heart regions,
   endothelial dysfunction, and cardiac impairment linked to apoptosis.
   Although considerable research has been dedicated to elucidating the
   mechanisms of ICM, our understanding remains insufficient [49][9].
   Genes and transcription factors play a key role in the mechanisms of
   ischemic heart disease. The process of coronary atherosclerosis is
   complex, with genetic diversity being a key factor. For instance,
   mutations in the low-density lipoprotein receptor (LDLR) gene could
   impair the clearance of LDL, resulting in elevated LDL levels and
   accelerated atherosclerosis [50][10]. Furthermore, genetic variants
   closely associated with physiological processes, including inflammatory
   responses, oxidative stress, and lipid metabolism, also contribute to
   the initiation and progression of atherosclerosis [51][11],
   [52][12].Vascular recanalization procedures have partially restored
   blood flow to the myocardium. However, reperfusion injury has become a
   significant challenge in managing ischemic cardiomyopathy. During the
   reperfusion phase, nuclear factor kappa B (NF-kappa B) acts as a
   crucial transcriptional regulator, demonstrating significant activation
   and influencing the expression of various downstream genes involved in
   myocardial cell apoptosis and survival [53][13]. The cytokine signal
   transducer and activator of transcription 3(STAT3) is essential for
   myocardial protection, as it regulates the synthesis of protective
   proteins and sustains the function of the mitochondrial electron
   transport chain (ETC). Its phosphorylated form inhibits the opening of
   the mitochondrial permeability transition pore (mPTP), thereby
   enhancing cardiomyocyte respiration and offering protection against
   ischemia/reperfusion (I/R) injury [54][14]. Hypoxia is crucial for
   regulating hypoxia-inducible factor (HIF). HIF-2 alpha helps protect
   the heart from injury by controlling the expression of amphiregulin
   (AREG) during ischemia/reperfusion [55][15].Jun and Fos, key components
   of the activator protein-1 (AP-1) complex, are involved in regulating
   inflammatory factor expression and participate in the I/R response
   [56][16]. Additionally, peroxisome proliferator-activated receptors
   (PPARs) are implicated in fatty acid metabolism and oxidative stress,
   serving to mitigate I/R injury and playing a vital role in protecting
   the myocardium during ischemic events [57][17].

   High-throughput technologies, such as biochips and next-generation
   sequencing, have produced large datasets that clarify the mechanisms
   behind the onset and progression of ICM. The ongoing discovery of
   specific biomarkers offers important insights into the mechanisms and
   potential treatments for ICM. For example, research by Yue Zheng et al.
   identified nine genes related to autophagy and ferroptosis, including
   IL-6 and prostaglandin-endoperoxide synthase 2(PTGS2), which might
   serve as promising therapeutic targets for patients with myocardial
   infarction [58][18]. In a similar vein, Dang et al. pinpointed genes
   such as small nuclear ribonucleoprotein polypeptides B (SNRPB), BLM
   RecQ like helicase(BLM), and ribosome biogenesis regulator 1
   homolog(RRS1), along with transcription factors like palladin,
   cytoskeletal associated protein(PALLD), thrombospondin 4(THBS4), and
   ATPase Na+/K + transporting subunit alpha 1(ATP1A1), as potential
   therapeutic targets for ICM [59][19]. These biomarkers are essential
   for the early identification of ICM, assessing risk, and monitoring
   disease progression. Therefore, discovering these biomarkers is a key
   focus of clinical research. Additionally, they have the potential to
   inform personalized treatment approaches and assess responses to
   therapy. However, the lack of significant and specific markers for ICM
   still poses challenges for its diagnosis and management. Consequently,
   there is an urgent necessity to investigate novel diagnostic markers
   associated with pathogenic mechanisms to formulate enhanced treatment
   strategies for ICM.

   Brent R. Stockwell first proposed the concept of “ferroptosis” in 2012
   to describe a unique type of cell death that depends on iron and
   reactive oxygen species (ROS). This distinguishes ferroptosis from
   other forms of cell death, including necroptosis, apoptosis,
   pyroptosis, and autophagy-related cell death [60][20]. Oxidative stress
   arises from an imbalance between ROS production and the ability of the
   cellular antioxidant system to neutralize these reactive species
   [61][21].Ferroptosis is significantly associated with the process of
   myocardial ischemia–reperfusion (I/R) injury and inhibiting ferroptosis
   has been demonstrated to alleviate such damage [62][22],
   [63][23].Additionally, ferroptosis is involved in the pathophysiology
   of heart failure and myocardial infarction, both of which could lead to
   the development of fibrosis. Therefore, ferroptosis might be pivotal in
   the etiology of cardiovascular diseases. The release of high levels of
   iron into coronary circulation could cause oxidative harm to the
   myocardium, while specific ferroptosis inhibitors might reduce ROS
   levels within the myocardium, thus improving cardiac function [64][24].
   Consequently, ferroptosis is thought to play a significant role in ICM
   caused by chronic ischemia and hypoxia. However, the exact molecular
   mechanisms are still unclear, emphasizing the urgent need to discover
   new therapeutic targets. Furthermore, the strategic induction or
   inhibition of ferroptosis offers a promising avenue for the treatment
   of ischemic diseases.

   This study analyzes the significance of ferroptosis-related
   differentially expressed genes (DEGs) in ICM. This study uses advanced
   bioinformatics methods and databases to identify ferroptosis-associated
   DEGs in samples from ICM patients. After identifying the DEGs, we will
   perform functional annotation and pathway enrichment analyses to better
   understand their biological roles and interactions. This systematic
   approach enables the discovery of hub genes that could play a crucial
   role in the development of ICM. This research provides new insights
   into the mechanisms of ICM and offers fresh perspectives for its
   clinical diagnosis and management.

2. Materials and methods

2.1. Data sources and processing

   The training sets for this study included two gene expression datasets
   related to ICM: [65]GSE57338, which uses the Affymetrix Human Gene 1.1
   ST Array [66][25], and [67]GSE5406, based on the Affymetrix Human
   Genome U133A Array [68][19]. The [69]GSE76701 dataset served as the
   validation set.All relevant clinical information was sourced from the
   Gene Expression Omnibus (GEO) database
   ([70]https://www.ncbi.nlm.nih.gov/geo). The analyzed samples came from
   humans, and their gene expression profiles were obtained using
   microarray technology.The [71]GSE57338 dataset included data from 95
   ICM patients and 136 controls, while [72]GSE5406 contained data from
   108 ICM patients and 16 controls ([73]Table 1). The raw sequencing data
   underwent comprehensive analysis. Box plots were used to illustrate the
   distribution of gene expression data. Both datasets represent
   myocardial samples from ischemic areas and have been previously
   employed in research pertaining to ischemic heart disease [74][26],
   [75][27], [76][28], [77][29].

Table 1.

   Identification of ferroptosis-related DEGs in ICM samples by GEO
   microarray data analysis.
   [78]GSE5406 [79]GSE57338
   Organism Homo sapiens Homo sapiens
   Experiment type Expression profiling by array Expression profiling by
   array
   Platforms [80]GPL96 [HG-U133A] [81]GPL11532 [HuGene-1_1-st]
   


   Sample(number)
    Normal 16 136
    Ischemic cardiomyopathy (ICM) 108 95
    Total 124 231
   [82]Open in a new tab

2.2. Identification of ferroptosis-related differentially expressed genes
(DEGs)

   To identify the biological differences between patients with ICM and
   control groups, we used the limma package in R to pinpoint DEGs.
   Because our datasets contained a limited number of differential genes,
   we broadened our screening criteria to include a log[2]-fold change
   (FC) greater than 0 for further analysis. A gene was considered
   differentially expressed if the absolute value of log[2]-FC was greater
   than 0 and the adjusted P-value was less than 0.05. To evaluate the
   expression of ferroptosis-related genes across all samples, we
   extracted those linked to iron-dependent cell death from the GeneCards
   database. In total, we retrieved 406 ferroptosis-related genes with a
   correlation score greater than 0.1 from GeneCards. An intersection
   analysis was then conducted to determine the genes that significantly
   overlapped with the DEGs identified in our investigation. The
   overlapping DEGs associated with ferroptosis were visually represented
   using Venn diagrams, and a heat map was created to display these
   findings. Additionally, boxplots illustrating the intersection of gene
   expression profiling data and ferroptosis-related genes for both ICM
   patients and control subjects were generated using the “ggpubr” package
   in R. Pearson’s correlations among the ferroptosis-related DEGs were
   calculated and visualized utilizing the “ggcorrplot” and “corrplot”
   packages in R.

2.3. Functional and pathway enrichment analysis for ferroptosis-related DEGs

   We analyzed the overlap between DEGs and ferroptosis-related genes
   using Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes
   (KEGG) enrichment assessments. GO analysis helps us understand the
   biological and molecular functions of genes. It provides functional
   annotations in categories such as biological processes (BP), molecular
   functions (MF), and cellular components (CC) [83][30]. In contrast,
   KEGG is a well-known database that provides insights into biological
   pathways and genomes related to diseases and drugs [84][31]. We used
   the R package ClusterProfiler [85][32]to perform functional annotations
   for all DEGs and identify significantly enriched biological processes.
   The outcomes of the enrichment analyses were illustrated using the R
   packages GOplot and topGO [86][33], with a significance threshold
   established at P < 0.05. We used gene set enrichment analysis (GSEA) to
   determine if a predefined set of genes showed significant variations
   between the two biological states. GSEA is commonly employed to
   evaluate changes in the activity of biological processes and pathways
   within gene expression datasets [87][34]. To assess the biological
   diversity between the two groups, gene sets such as
   “c5.go.v7.4.entrez.gmt” and “c2.cp.kegg.v7.4.entrez.gmt” were obtained
   from the MSigDB website [88][35],and the R package “clusterProfiler”
   (v3.4.4) was applied for executing and visualizing GSEA. Additionally,
   gene set variation analysis (GSVA) [89][36], an unsupervised and
   non-parametric technique for gene set enrichment, was employed to
   convert a gene matrix into a gene set using various sample matrices,
   thereby estimating variations in biological pathways based on GSEA
   results derived from microarray transcriptome data. The
   “c2.cp.kegg.v7.4.entrez.gmt” gene sets were utilized to compute
   single-sample gene set enrichment scores. Differential pathway analysis
   was carried out using the R package “limma” [90][37] and the results
   were visualized through the R package “pheatmap.” A significance level
   of P < 0.05 was established.

2.4. Construction of protein–protein and TF–hub-gene interaction networks

   To clarify the connections between genes involved in ferroptosis, we
   established a protein–protein interaction (PPI) network using the
   Search Tools for the Retrieval of Interacting Genes (STRING) with the
   previously mentioned 53 genes as input [91][32]. We set the confidence
   threshold to the default level of 0.4. Next, we exported the network
   and visualized it using Cytoscape [92][33]for further analysis.
   Parameters for each node within the network were calculated, and we
   explored the co-expression network nodes (hub genes) using the
   cytoHubba plugin, applying a standard reference of K score = 2, a
   cut-off degree of 2, and a cut-off node score of 0.2 [93][34]. These
   nodes are highly interconnected, indicating their potential importance
   in regulating essential biological processes. This aspect is crucial
   for ongoing investigations. We represented the expression levels of 11
   hub genes, obtained from two genome-wide expression profiling datasets,
   as box plots. Additionally, transcription factors (TFs) were identified
   based on the ChIP-X Enrichment Analysis 3 (ChEA3) database
   corresponding to the hub gene nodes [94][35]. The top ten TFs were
   subsequently imported into Cytoscape for further analysis and
   visualization, which included network diagrams illustrating
   interactions between hub genes and TFs, network diagrams among TFs, bar
   graphs depicting TF-target gene coverage and TF-scores, as well as heat
   maps representing correlation analyses between TFs and hub genes.

2.5. Construction of small molecular drug–hub-gene interaction network

   To clarify the association between hub genes and sensitivity to
   small-molecule drugs, we retrieved drug IC50 values from the Cancer
   Cell Line Encyclopedia (CCLE) [95][36] and the Genomics of Drug
   Sensitivity in Cancer (GDSC) databases. We integrated this data with
   information from the RNAactDrug database [96][38]. After integrating
   the data, we independently assessed the relationships between hub genes
   and drug IC50 values, employing bubble diagrams to analyze each
   database.

2.6. Hub gene–gene correlations and interaction networks

   To focus on the biologically relevant correlations among hub genes, the
   R package Corrplot was utilized to create correlation plots. The
   interaction relationships among hub genes were calculated using STRING,
   and interaction networks were constructed using Cytoscape.

3. Results

3.1. The relationship between ferroptosis and DEGs

   [97]Fig. 1 shows the steps we followed to identify and select relevant
   studies. Due to significant batch effects in datasets from different
   sources, we evaluated the gene expression distribution in the raw data
   before and after batch correction (see [98]Supplementary Fig. S1).
   Before batch correction, the [99]GSE57338 dataset had 4479 upregulated
   genes and 4943 downregulated genes, while the [100]GSE5406 dataset had
   700 upregulated genes and 859 downregulated genes. These results
   highlighted the significant impact of batch effects when the data were
   combined without correction. Each sample set from various sources
   exhibited unique distribution patterns and expression levels.
   Therefore, we used the normalizeBetweenArrays function from the limma
   package to standardize the variable data, keeping in mind that this
   function is intended for use within a single dataset. Following the
   correction of batch effects and the application of log normalization,
   the data exhibited a convergence in overall expression distributions
   across all samples, thereby enhancing the robustness and precision of
   the subsequent analyses.

Fig. 1.

   [101]Fig. 1
   [102]Open in a new tab

   The study flow chart.DEGs: differentially expressed genes; PPI:
   protein–protein interation; GO: Gene Ontology; KEGG: Kyoto Encyclopedia
   of Genes and Genomes; GSEA: Gene set enrichment analysis; GSVA: Gene
   set variation analysis; TFs: transcription factors.

   [103]Fig. 2 illustrates the comparison of gene expression between
   normal and ICM samples using heat maps and volcano plots. The
   [104]GSE5406 dataset revealed that 859 genes had reduced expression,
   while 645 genes showed increased expression (see [105]Fig. 2(a and b),
   adj. P < 0.05). In the [106]GSE57338 dataset, 4605 genes were
   upregulated, while 5059 genes were downregulated (see [107]Fig. 2(c and
   d), adj.P < 0.05). Among these genes, 53 were classified as
   ferroptosis-related differential genes, identified by intersecting the
   differentially expressed genes with those known to be associated with
   ferroptosis. [108]Supplementary Figs. S2 and S3, along with [109]Table
   S1, show the distribution variations of these ferroptosis-related
   differential genes across the two datasets. The expression heat maps
   for the 53 identified ferroptosis-related differential genes are
   displayed in [110]Fig. 3(a and b), while the Venn diagram in [111]Fig.
   3(c) illustrates the overlap among the three datasets.

Fig. 2.

   [112]Fig. 2
   [113]Open in a new tab

   Volcano plots and heat maps of differential gene expression. (a) The
   heat map displays the differential gene expression from the
   [114]GSE5406 dataset. Each row corresponds to a different gene, while
   each column represents an individual sample, with the color intensity
   indicating gene expression levels. (b) The volcano plot for the
   [115]GSE5406 dataset visually represents the same differential gene
   expression data, plotting the log[2] fold change (Log[2] FC) on the
   x-axis against the statistical significance (−log[10] q-value) on the
   y-axis. In this plot, red dots represent significantly upregulated
   genes (645 in total), blue dots indicate downregulated genes (895), and
   grey dots signify genes with no significant change (P < 0.05). (c) The
   heat map for the [116]GSE57338 dataset similarly illustrates the
   differential gene expression across samples. (d) The corresponding
   volcano plot for the [117]GSE57338 dataset reveals an extensive number
   of differential genes, showing a total of 4,605 upregulated and 5,095
   downregulated genes.

Fig. 3.

   [118]Fig. 3
   [119]Open in a new tab

   The expression of ferroptosis-related differential genes in
   [120]GSE5406 and [121]GSE57338 and Venn diagram of the intersection of
   differentially expressed genes and ferroptosis-related differential
   genes. (a) The heatmap illustrates the expression levels of
   ferroptosis-related differential genes in the [122]GSE5406 dataset.
   Each row corresponds to a specific gene, while each column represents
   an individual sample. The color gradient indicates expression levels,
   with red dots denoting high expression and blue dots indicating low
   expression. (b) The heatmap for the [123]GSE57338 dataset presents the
   expression of the ferroptosis-related differential genes. Each row
   represents a gene, and each column represents a sample. High expression
   is represented by red, whereas low expression is represented by blue.
   (c) The Venn diagram shows the intersection of differentially expressed
   genes and ferroptosis-related genes. Following the intersection of the
   406 differential genes with ferroptosis-related genes, a total of 53
   genes were identified as ferroptosis-related differential genes.

3.2. Correlation analysis

   To further clarify the expression patterns of ferroptosis-related
   genes, box plots were created for the 53 differential genes across
   various datasets. The statistical analysis showed that all 53 genes had
   significant P-values (P < 0.05), except for chaperonin containing TCP1
   subunit 4(CCT4). Spearman correlation analysis was performed to assess
   the relationships among these genes, revealing significant correlations
   for all 53 ferroptosis-related differential genes (P < 0.05). [124]Fig.
   4(a and b) show the correlation heat maps for the two datasets.

Fig. 4.

   [125]Fig. 4
   [126]Open in a new tab

   Correlation analysis of ferroptosis-related differential genes. (a) The
   correlation heatmap for the [127]GSE5406 dataset illustrates the
   relationships among the 53 identified ferroptosis-related differential
   genes. A color code is employed to denote the strength of correlation,
   with blue indicating a strong positive correlation (correlation
   coefficient close to 1.0) and red indicating a strong negative
   correlation (correlation coefficient close to −1.0). The intensity of
   the color change reflects the degree of linear correlation, confirming
   that all correlation values were significant (P < 0.05). (b) The
   correlation heatmap for the [128]GSE57338 dataset presents the
   interactions among the same set of ferroptosis-related differential
   genes. The heatmap visually represents how these genes correlate with
   one another, with all correlations remaining significant (P < 0.05).

3.3. GO/KEGG enrichment analysis

   To evaluate the biological processes and functions affected by the DEGs
   related to ferroptosis, we performed GO, KEGG, and GSEA as shown in
   [129]Fig. 5, [130]Fig. 6. We identified 247 significant GO terms across
   the BP, CC, and MF categories (see [131]Supplementary Table S2).
   [132]Fig. 5(a and b) visually represent the twenty most significant GO
   terms using bubble and bar graphs. Several highly ranked BP terms,
   including “cellular response to chemical stress”, “cellular response to
   oxidative stress”, “response to oxidative stress”, “cellular oxidant
   detoxification” and “response to toxic substance”, suggest a novel role
   for ferroptosis-related DEGs in the oxidative stress mechanisms linked
   to ICM development.

Fig. 5.

   [133]Fig. 5
   [134]Open in a new tab

   GO/KEGG enrichment analysis of ferroptosis-related differential genes.
   (a) The bar chart illustrates the results of the Gene Ontology (GO)
   enrichment analysis for ferroptosis-related differential genes. The
   x-axis displays the negative logarithm of P-value (−log10), indicating
   the statistical significance of each enriched term, while the y-axis
   presents the top 20 GO terms categorized into biological processes
   (BP), molecular functions (MF), and cellular components (CC). The color
   of the bars reflects the corrected P-value, highlighting the most
   significantly enriched terms related to oxidative stress response.
   (b)The bubble diagram visualizes the GO enrichment outcomes. The
   proportion of enriched genes is depicted on the x-axis, while the GO
   terms are plotted on the y-axis. (c) The bar chart for the KEGG pathway
   enrichment analysis presents the top 40 pathways significantly
   associated with the ferroptosis-related differential genes (P < 0.05).
   The x-axis reflects the negative logarithm of P-value (−log[10]), with
   each pathway illustrated on the y-axis. Key pathways such as the TNF
   signaling pathway and NOD-like receptor signaling pathway are
   prominently represented, indicating their relevance to the immune
   response. (d)The bubble diagram displays the KEGG enrichment results.
   The proportions of the enriched genes are shown on the x-axis, while
   the KEGG pathways are plotted on the y-axis. Each bubble's size
   indicates the gene count for each pathway, and the color illustrates
   the adjusted P-value, highlighting pathways that are notably influenced
   by differential gene expression during ferroptosis. GO, Gene Ontology;
   KEGG, Kyoto Encyclopedia of Genes and Genomes.

Fig. 6.

   [135]Fig. 6
   [136]Open in a new tab

   GSEA/GSVA enrichment analysis. (a) The heat map showcases the GSVA
   enrichment pathways in the [137]GSE5406 dataset. The x-axis represents
   the Log[2] (Fold change, FC) of pathway activity, while the y-axis
   displays the −log[10] (P-value) for statistical significance. Each
   colored node corresponds to a specific pathway, with upregulated
   pathways indicated by red, downregulated pathways shown in blue, and
   unchanged pathways represented by grey spots. (b) The bar graph of the
   [138]GSE5406 dataset illustrates the GSVA enrichment pathways in
   [139]GSE5406, where the x-axis indicates the enrichment score for each
   KEGG pathway, and the y-axis lists the names of these pathways. (c) The
   heat map in [140]GSE57338 depicts GSVA enrichment pathways, where the
   x-axis shows the Log[2] (Fold change, FC), and the y-axis shows the
   −log[10] (P-value). Each node is occupied by a pathway. Upregulated,
   downregulated, and unchanged pathways are indicated by red, blue, and
   grey spots, respectively. (d) The corresponding bar graph for
   [141]GSE57338 presents enrichment scores along the x-axis and KEGG
   pathway names along the y-axis. (e) The GSEA results for [142]GSE5406
   are visualized, with the x-axis showing the rank of differentially
   expressed genes (where > 0 indicates upregulation and < 0 indicates
   downregulation). The upper y-axis denotes the enrichment score, while
   the lower y-axis reflects the logFC values. Each color represents a
   different enriched pathway. (f) The GSEA results for [143]GSE57338 are
   presented, emphasizing the differential expression of genes across
   various pathways and their significance. (g) Ridgeline plots for GSEA
   results in [144]GSE5406 depict the expression distribution of
   core-enriched genes across GSEA enrichment categories. The colors of
   the ridges correspond to the p-values, while the horizontal axis
   indicates the probability distribution, elucidating the impact of
   pathways on gene expression. (h) The ridgeline plots for [145]GSE57338
   providing insights into the expression distributions of core genes and
   their respective enrichment categories. GSEA, Gene set enrichment
   analysis; GSVA, Gene set variation analysis; KEGG, Kyoto Encyclopedia
   of Genes and Genomes.

   The changes observed in the CC category indicate that components such
   as “cytoplasmic region”, “chaperonin complex”, “chaperonin-containing T
   complex”, “chaperone complex”,“melanosome pigment” and “granule” might
   contribute to the cellular changes associated with ferroptosis. The MF
   analysis highlighted the involvement of “antioxidant” and
   “oxidoreductase activities” in biological processes related to ICM.
   Furthermore, the KEGG pathway enrichment analysis revealed 60
   significant pathways (P < 0.05). The bar charts and bubble diagrams
   illustrating the top 40 ranked KEGG pathways associated with DEGs are
   presented in [146]Fig. 5(c and d), which include immune response
   pathways such as the TNF signaling pathway, NOD-like receptor signaling
   pathway, acute myeloid leukemia, human cytomegalovirus
   infection,shigellosis, apoptosis,hepatocellular carcinoma, chemical
   carcinogenesis receptor activation, chemical carcinogenesis reactive
   oxygen species and coronavirus disease COVID-19.

   Additionally, the network graphs depicting the GO/KEGG enrichment
   pathway categories could be found in [147]Supplementary Fig. S4(a and
   b), while the intricate relationships between ferroptosis-related DEGs
   and the GO/KEGG enrichment pathways are detailed in [148]Supplementary
   Fig. S4(c and d). The comprehensive GO/KEGG pathway networks are
   illustrated in [149]Supplementary Fig. S4(e and f).

3.4. Gene set enrichment analysis (GSEA) and gene set variation analysis
(GSVA)

   To further elucidate the functional relationships among DEGs associated
   with ferroptosis, we employed GSEA and GSVA methodologies across two
   microarray datasets. Detailed results for each dataset are presented in
   [150]Supplementary Table S3. A total of 51 pathways exhibiting
   significant enrichment were identified within the GO pathway gene sets.
   The line graphs representing the most significantly enriched pathways
   are illustrated in [151]Fig. 6(e and f), while the volcanic plots for
   the top five pathways are presented in [152]Fig. 6(g and h). The
   pathways 'oxidoreductase activity' and 'response to oxidative stress'
   were identified as the most significant. Their strong association with
   oxidative stress regulation reinforces the findings from the GO
   enrichment analysis.

   The GSVA approach enabled the transformation of KEGG pathway
   annotations into new expression matrices for the two gene sets. In the
   [153]GSE5406 dataset, 673 KEGG pathways showed significant enrichment.
   The three most prominent pathways related to disease are 'response to
   the mechanistic target of rapamycin', 'head and neck cancers' and
   'targets of Elvidge HIF1A and HIF2A' (see [154]Fig. 6(a and b)). In the
   [155]GSE57338 dataset, 2215 KEGG pathways were significantly enriched.
   The three most notable pathways are kaab failed heart ventricle dn,
   interaction of Martin with histone deacetylases and kaab failed heart
   atrium dn. These pathways have been associated with cardiac diseases
   and pharmacological effects, which correspond with the clinical
   conditions of the patients represented in the dataset (see [156]Fig.
   6(c and d)).

3.5. Construction of PPI networks

   We investigated the interactions of genes related to ferroptosis by
   reconstructing and visualizing the predicted PPI network of hub genes
   using the cytoHubba plug-in in Cytoscape, as shown in
   [157]Supplementary Fig. S5(a – c). In networks, nodes with a higher
   out-degree are more likely to connect to nodes with a lower in-degree,
   which significantly affects the overall network dynamics. Through
   genome mining, we identified 11 hub genes: chaperonin containing TCP1
   subunit 4(AHCY), chaperonin containing TCP1 subunit 3(CCT3), chaperonin
   containing TCP1 subunit 4(CCT4), chaperonin containing TCP1 subunit
   5(CCT5), chaperonin containing TCP1 subunit 6A(CCT6A), eukaryotic
   translation initiation factor 4A (EIF4A), enolase 1(ENO1), nascent
   polypeptide associated complex subunit alpha(NACA), ribosomal protein
   lateral stalk subunit P0(RPLP0), ribosomal protein S2(RPS2), and
   ribosomal protein S7(RPS7). We specifically focused our validation
   efforts on eight of these hub genes: AHCY, CCT4, CCT5, CCT6A, ENO1,
   NACA, RPLP0, and RPS7. We assessed the significance of the DEGs using
   the Wilcoxon Rank-Sum test. The results showed that all hub genes,
   except CCT4 (P-value = 0.06), achieved statistical significance with
   P-values less than 0.05 (see [158]Fig. 7(a and b)).

Fig. 7.

   [159]Fig. 7
   [160]Open in a new tab

   Hub gene analysis. To delve deeper into the interactions among the
   differentially expressed iron death genes, we constructed a
   protein–protein interaction (PPI) network using the STRING database and
   visualized the results with Cytoscape software. (a) The boxplot
   illustrates the expression levels of hub genes identified within the
   [161]GSE5406 dataset. The x-axis represents specific genes, while the
   y-axis displays their corresponding expression values across samples.
   Different colors denote distinct sample groups, allowing for a visual
   comparison of gene expression. The midline in each box reflects the
   median expression value, with the upper and lower bounds representing
   the upper and lower quartiles, respectively. This visualization helps
   to elucidate the distribution and variability of hub gene expression
   within this dataset. (b) The boxplot in the [162]GSE57338 dataset
   provides a comparative view of the same hub genes. As with panel (a),
   the axes represent gene identity and expression values, with color
   coding indicating sample groups. The statistical representation
   highlights the expression range and central tendency of each gene,
   facilitating an understanding of their potential roles in the
   biological processes associated with ferroptosis.

3.6. The construction of TF–hub-gene interaction networks

   We identified TFs associated with specific target genes through the use
   of ChEA3 libraries. [163]Supplementary Fig. S6 shows the top ten
   transcription factors ranked by their hub scores. [164]Supplementary
   Fig. S7(a and b) provide a clear representation of the regulatory
   networks and genetic contexts associated with the two datasets.
   Additionally, [165]Supplementary Fig. S7(c) presents the interaction
   networks for each transcription factor. The results indicate that these
   transcription factors, commonly found among co-expressed genes, might
   have similar biological roles and engage in related regulatory
   processes.

3.7. Drug sensitivity analysis

   To clarify the relationship between the eleven hub genes and
   sensitivity to small-molecule therapeutics, we integrated data from
   three drug sensitivity repositories: CCLE and GDSC. The correlation
   bubble plots showing the relationship between gene expression levels
   and IC50 values of small-molecule drugs are presented in
   [166]Supplementary Fig. S8(a and b). An increased IC50 value indicates
   greater drug resistance. Therefore, a negative correlation suggests
   that higher gene expression is associated with increased drug
   sensitivity, while a positive correlation indicates the opposite. Our
   analysis revealed that NACA had a negative correlation with most
   small-molecule drugs in both datasets. This finding suggests that
   changes in NACA expression might be linked to drug sensitivity,
   indicating the need for further investigation of this gene for
   potential clinical insights.

3.8. Hub gene–gene correlations and interaction networks

   The hub genes tend to come from the same gene family, indicating strong
   interrelations and significant associations among them. Consequently,
   we calculated the biologically relevant correlations among the 11 hub
   genes. These are illustrated in [167]Fig. 8(a and b). Most of these
   gene-to-gene correlations were positive, with only a few being
   negative. In the [168]GSE5406 dataset, the gene pairs with the highest
   positive correlation were 'EIF4A1-ENO1′ and 'RPLP0-RPS2′,both having a
   correlation coefficient of 0.83, as shown in [169]Fig. 8(a). Likewise,
   in the [170]GSE57338 dataset, the gene pair with the highest positive
   correlation was 'RPLP0-RPS2′,which attained a correlation coefficient
   of 0.87, as represented in [171]Fig. 8(b). The Gene Ontology (GO)
   pathways, adjusted for interaction enrichment, are displayed in
   [172]Fig. 8(c). The top three identified pathways are
   'chaperonin-containing t-complex', 'positive regulation of protein
   localization to telomere', and 'positive regulation of protein
   localization to Cajal body'. All these pathways are related to protein
   localization. This observation strongly suggests the presence of
   reciprocal regulatory interactions between the principal disease genes
   linked to ischemic cardiomyopathy and the mechanisms that regulate
   protein localization.

Fig. 8.

   [173]Fig. 8
   [174]Open in a new tab

   Hub-gene interaction network analysis. (a) The heatmap displays the
   correlation analysis of microarray data from the [175]GSE5406 dataset,
   revealing the relationships among the 11 hub genes. Each square
   represents the correlation coefficient between two genes, with blue
   indicating values close to 1.0 (strong positive correlation) and red
   indicating values close to −1.0 (strong negative correlation). Notably,
   the most significant positive correlations were observed between EIF4A1
   and ENO1, as well as between RPLP0 and RPS2, with a correlation
   coefficient of 0.83. (b) The heatmap for the [176]GSE57338 dataset
   demonstrates the correlation analysis among the 11 hub genes. Here,
   RPLP0 and RPS2 exhibit the highest correlation coefficient at 0.87,
   underscoring the strong positive relationship among these hub genes.
   (c) The Gene Ontology (GO) interaction enrichment map for the 11 hub
   genes illustrates the biological processes in which they are involved.
   The x-axis indicates the proportion of genes associated with each GO
   term, while the y-axis lists the respective GO terms. The color
   gradient conveys false discovery rate (FDR) values, highlighting the
   top three pathways related to protein localization processes, which
   suggests potential interactions of key genes in myocardial diseases
   regarding protein localization. GO, Gene Ontology; FDR, false discovery
   rate.

4. Discussion

   The worldwide incidence and mortality of ischemic heart disease (IHD)
   remains significant, with over 7 million deaths each year, representing
   approximately 13 % of all global deaths, as reported by the WHO
   [177][38].This concerning situation is mainly linked to the progression
   of coronary atherosclerosis, which narrows blood vessels and could lead
   to complete blockage, significantly reducing blood flow to the heart
   muscle [178][5].The changes associated with this condition often lead
   to serious complications, such as angina, recurrent episodes, heart
   attacks, heart failure, and sudden cardiac death. Although there have
   been significant advancements in medical technology that have improved
   the diagnosis and management of ischemic heart disease, including
   better pharmacotherapy, minimally invasive techniques, and advanced
   surgical procedures, current treatment strategies still encounter many
   challenges. These challenges are especially clear in early diagnosis
   and accurate disease prediction, indicating a significant opportunity
   for further improvements in these critical areas [179][39].Our primary
   objective is to pinpoint specific biomarkers related to ICM through
   careful filtering methods and to develop a detailed map of disease
   markers based on molecular networks. Bioinformatics is essential for
   meeting this need and achieving our goals. Through detailed analysis of
   large biological datasets, bioinformatics clarifies the disease's
   underlying mechanisms and provides valuable insights into essential
   molecular processes. Consequently, we focus our study on
   molecular-level investigations of ischemic heart disease using advanced
   bioinformatics techniques. This initiative enhances our understanding
   of ICM's pathophysiology. It also lays a solid theoretical foundation
   and provides crucial support for developing innovative diagnostic tools
   and clinical treatments.

   Over the past decade, mechanisms of cell death, including apoptosis and
   necrosis, have been identified as significant contributors to ICM and
   myocardial fibrosis [180][40], [181][41].Ferroptosis was first
   identified as a caspase-independent pathway of cell death, primarily
   triggered by the depletion of glutathione (GSH) due to the inhibitor
   erastin. The reduction of GSH undermines the activity of GSH peroxidase
   4 (GPX4), which depends on GSH as a reducing co-substrate, thus
   hindering the detoxification of lipid hydroperoxides (LOOH) and
   obstructing their conversion back to the corresponding lipid alcohols
   [182][42]. Additionally, ferroptosis could be instigated by
   pharmacological inhibition of GPX4 or through the targeted knockout of
   the GPX4 gene. Changes in lipid metabolism that increase the
   availability of oxidizable polyunsaturated lipids or disrupt iron
   homeostasis might also heighten cellular susceptibility to
   ferroptosis.Lipid peroxidation is a crucial factor in the execution
   phase of ferroptosis, while iron is vital for the accumulation of lipid
   peroxides, influencing cell sensitivity to this type of death
   [183][43]. Numerous agents, regulators, and inhibitors of ferroptosis
   influence ROS accumulation in an iron-dependent manner
   [184][44].Erastin, a small molecule, was the first known inducer of
   ferroptosis, functioning through ROS- and iron-dependent signaling
   pathways [185][20]. Ectonucleotide pyrophosphatase/phosphodiesterase
   2(ENPP2) is a crucial protein that serves as a lipid phosphatase,
   playing an essential role in preventing ferroptosis induced by erastin
   [186][45]. Nonetheless, the exact mechanisms through which ferroptosis
   contributes to ICM remain insufficiently elucidated, highlighting the
   need for further exploration of its biological functions and related
   signaling pathways.

   Researchers are increasingly focusing on the relationship between
   ischemic injury and ferroptosis, particularly in the context of
   cardiovascular diseases and their complications. There may be a
   temporal and bidirectional relationship between cell damage resulting
   from ischemic events, such as myocardial ischemia and myocardial
   infarction, and ferroptosis, warranting in-depth exploration of this
   complex mechanism [187][46]. Firstly, from a temporal perspective, the
   correlation between ischemic injury and ferroptosis manifests in
   several ways. Following ischemia, there are drastic changes in the
   intracellular environment, including hypoxia and glucose deprivation,
   which lead to metabolic disturbances and increased production of ROS.
   Early studies have indicated that shortly after ischemia, cells may be
   induced to undergo ferroptosis due to elevated oxidative stress
   [188][47]. Excessive ROS triggers lipid peroxidation, which compromises
   the integrity of the cell membrane and ultimately precipitates the
   occurrence of ferroptosis. In fact, markers of ferroptosis may change
   within hours of ischemia, indicating a significant temporal correlation
   between ischemia-induced ferroptosis [189][48]. However, the temporal
   aspect is not limited to the occurrence of ferroptosis following
   ischemia. Ferroptosis not only responds to ischemic injury but may also
   exacerbate subsequent cellular damage. This interplay can lead to
   aggravated injury; for instance, cell death caused by ferroptosis would
   result in decreased functional capacity of heart tissue, thereby
   increasing the risk of heart failure. In this context, ferroptosis can
   be viewed both as an outcome of ischemic injury and as a factor that
   promotes the progression of ischemic heart disease. Therefore, there
   exists a bidirectional relationship between ischemic injury and
   ferroptosis, wherein each process may amplify the negative effects of
   the other [190][49].Research on the bidirectionality of this
   interaction is also emerging. Ferroptosis can induce persistent
   oxidative stress, causing damage to endothelial cells and subsequently
   leading to endothelial dysfunction, which facilitates the formation of
   atherosclerosis. As atherosclerosis progresses, the heart's ability to
   adapt to ischemic events diminishes, creating a vicious cycle.
   Additionally, whether ischemic events trigger ferroptosis is a question
   worth further investigation. For instance, in the context of acute
   myocardial infarction, ischemic conditions may lead to the release of
   intracellular iron ions, thereby activating iron-related cell death
   pathways and accelerating myocardial damage. Thus, there is a clear
   interplay between ischemic events and ferroptosis [191][50].

   In ICM, hypoxia is a prominent feature characterized by insufficient
   oxygen supply to myocardial tissues, which leads to numerous metabolic
   alterations. Under ischemic conditions, myocardial cells experience
   limited oxidative phosphorylation, prompting a metabolic shift toward
   anaerobic pathways. This switch precipitates lactic acid accumulation
   and intracellular acidosis, which are potential triggers for
   ferroptosis [192][51].Ferroptosis is characterized by the accumulation
   of intracellular lipid peroxidation products and ROS production through
   iron-dependent processes. This leads to lipid peroxidation, resulting
   in cellular membrane damage and cell death [193][20]. Ferroptosis is
   also associated with the excessive production of ROS due to oxidative
   stress [194][52]. Notably, hypoxia can alter iron acquisition and
   metabolism in cardiac cells, resulting in an increase in free iron
   levels. The elevated free iron can exacerbate lipid peroxidation
   reactions, making ferroptosis a critical event in the pathophysiology
   of ICM [195][53]. Furthermore, the inflammatory response is an
   essential pathogenic component in ICM. It activates a cascade of
   inflammatory signaling pathways, driven by pro-inflammatory cytokines
   such as tumor necrosis factor-alpha (TNF-α) and interleukin-6 (IL-6),
   which are known to be elevated in ICM cases. These cytokines not only
   induce apoptosis of cardiac myocytes but also contribute to oxidative
   stress and lipid peroxidation, thereby enhancing the susceptibility of
   myocardial cells to ferroptosis [196][54]. The infiltration of immune
   cells, particularly macrophages, during inflammatory states exacerbates
   oxidative stress by releasing ROS, hence increasing the vulnerability
   of cardiomyocytes to ferroptosis [197][55].The interplay between
   hypoxia and inflammation is critically interlinked in the progression
   of ICM. Hypoxia-induced stress can stimulate myocardial cells to
   release pro-inflammatory factors, which further intensifies the
   inflammatory response within the cardiac milieu [198][56]. Conversely,
   the resultant inflammatory milieu can exacerbate hypoxic injury by
   accelerating the damage to myocardial cells, leading to even more
   profound hypoxia and thereby establishing a vicious cycle. This
   interaction plays a pivotal role in the deterioration of cardiac
   function and highlights the complexity of managing ischemic heart
   disease [199][57].

   In our study, we utilized both the [200]GSE57338 and [201]GSE5406
   datasets. Many previous studies have established different sequencing
   microarrays specifically targeting ICM. For example, Chen et al.
   investigated DEGs related to ICM using the [202]GSE5406 and
   [203]GSE57338 microarray datasets. They found a significant association
   between myosin heavy chain 6(MYH6) expression and both ICM and heart
   failure (HF) [204][28]. Similarly, Zhang et al. identified DEGs
   associated with ICM and dilated cardiomyopathy (DCM) across five
   datasets: [205]GSE1869, [206]GSE5406, [207]GSE57338, [208]GSE79962, and
   [209]GSE116250, concluding that the ASPN gene could serve as a
   biomarker for heart failure [210][29]. Cao et al. also extracted
   ICM-related DEGs from the [211]GSE116250, [212]GSE46224, and
   [213]GSE5406 datasets. They identified five DEGs (SERPINA3, FCN3,
   COL3A1, HBB, and MXRA5) involved in the pathogenesis of heart failure
   associated with ICM [214][58]. Additionally, another investigation
   employed the [215]GSE5406 dataset for differential expression analysis,
   revealing 15 DEGs associated with ferroptosis in ICM, indicating that
   ferroptosis might contribute to the pathogenesis of ICM [216][59].
   Given the variability among datasets and the limitations of small
   sample sizes, it is crucial to validate these findings from various
   perspectives. As a result, we have formulated a novel signature of
   ferroptosis-related DEGs pertinent to ICM.

   Eleven DEGs are identified as key genes related to ferroptosis: AHCY,
   CCT3, CCT4, CCT5, CCT6A, EIF4A1, ENO1, NACA, RPLP0, RPS2, and RPS7.
   These genes have a complicated relationship with ICM, influencing its
   development in various ways. Recent research indicates that these genes
   might affect the progression of ICM. They do this through their roles
   in protein folding and assembly, cellular stress responses, and iron
   metabolism [217][60], [218][61]. The CCT (Chaperonin Containing TCP-1)
   gene family, which includes CCT3, CCT4, CCT5, and CCT6A, is recognized
   as a key factor in the development of ICM. These molecular chaperones
   are critical for the correct folding of cytoskeletal and other client
   proteins essential for cellular survival and functionality under stress
   conditions. Their role in ferroptosis shows that they help with proper
   protein folding and also influence cell death pathways.

   The CCT gene family encodes the subunits of the chaperonin-containing
   TCP1 complex, which plays a pivotal role in the proper folding of
   cytoskeletal proteins, including actin and tubulin. These proteins are
   essential for preserving the structural integrity and contractile
   capabilities of cardiomyocytes. In the context of ischemic
   cardiomyopathy, the accurate folding and assembly of these proteins
   become increasingly vital due to heightened cellular stress and the
   necessity for repair mechanisms. Notably, CCT3 has been implicated in
   both the folding of newly synthesized proteins and the refolding of
   misfolded proteins, a process that is particularly critical under
   ischemic conditions where protein damage is prevalent [219][62].In a
   similar vein, CCT4 and CCT5 are fundamental to the functionality of the
   chaperonin complex, ensuring that actin and tubulin retain their
   correct conformation and operational capacity [220][63]. Additionally,
   CCT6A, another subunit of this complex, contributes to the folding of
   cytoskeletal proteins and might hold particular significance in the
   context of cardiac stress responses [221][64].

   The dysfunction of any of these CCT subunits could result in
   compromised cytoskeletal integrity, thereby adversely affecting cardiac
   function. Research has indicated that mutations or altered expression
   of these genes might play a role in the pathogenesis of ischemic
   cardiomyopathy by disrupting the normal folding and functionality of
   critical structural proteins within the heart [222][65].Furthermore,
   CCT genes have been associated with oxidative stress and ferroptosis.
   For instance, CCT3, CCT4, CCT5, and CCT6A encode chaperone proteins
   that are subunits of the CCT. CCT3 has been shown to enhance cell
   proliferation by inhibiting ferroptosis and activating the AKT
   signaling pathway [223][66]. Following the inhibition of CCT3, the cell
   cycle is arrested in the G0/G1 phase, leading to impaired matrix
   metalloproteinases (MMP) activity and elevated intracellular ROS
   levels. Consequently, miRNA-mediated inhibition of CCT3 promotes
   apoptosis by modulating the homeostasis of intracellular ROS and the
   distribution of free amino acids involved in energy metabolism
   [224][67]. In our investigation, we observed that the expression levels
   of CCT were significantly lower in the ICM group compared to the
   healthy cohort. Further experimental validation is necessary to
   elucidate the functional relevance of CCT in ferroptosis and its
   association with the pathogenesis of ICM.

   Genes such as AHCY, EIF4A1, ENO1, NACA, RPLP0, RPS2, and RPS7 play
   essential roles in various cellular functions, including methylation,
   protein folding, translation initiation, glycolysis, and ribosomal
   activity. Additionally, researchers have studied these genes for their
   roles in cellular metabolism and stress responses in ICM [225][68].
   Disruptions in methylation could alter gene expression and are linked
   to several cardiovascular conditions, including ICM. Notably, elevated
   homocysteine levels due to AHCY dysfunction contribute to endothelial
   dysfunction and atherogenesis, both critical factors in the
   pathogenesis of ICM [226][64], During ischemia, regulating protein
   synthesis is crucial for cell survival. Changes in the expression or
   functionality of EIF4A1 could significantly impact the translation of
   essential proteins that govern cell survival and apoptosis, thereby
   influencing the progression of ICM [227][5]. ENO1 functions as a
   glycolytic enzyme that is crucial for energy production. Under ischemic
   conditions, the heart increasingly depends on anaerobic glycolysis to
   generate ATP.

   NACA is essential for the co-translational folding of nascent
   polypeptides and for directing proteins to the endoplasmic reticulum.
   Proper protein folding and trafficking are crucial for cellular
   function. Disruptions in these processes could lead to endoplasmic
   reticulum stress and apoptosis, which significantly contribute to the
   development of ICM [228][6]. ENO1 is also involved in several roles
   unrelated to glycolysis, such as binding to plasminogen and modulating
   immune responses. The dysregulation of ENO1 could disrupt energy
   metabolism and exacerbate the inflammatory response associated with ICM
   [229][69]. Zhang et al. found that ENO1 could inhibit mitochondrial
   iron-induced ferroptosis by reducing the expression of mitochondrial
   ferritin-1 [230][70]. RPLP0, RPS2, and RPS7 are integral components of
   the ribosome, which is essential for protein synthesis. Under
   conditions of ischemic stress, the regulation of protein synthesis
   becomes critical for cellular survival. Changes in the expression or
   functionality of ribosomal proteins could significantly influence the
   synthesis of proteins that are crucial for cell survival, apoptosis,
   and repair mechanisms, thereby affecting the progression of ICM
   [231][7]. Furthermore, additional bioinformatics investigations have
   corroborated the potential of ENO1 and RPS7 genes as diagnostic markers
   for ferroptosis. For instance, He et al. identified datasets of
   ferroptosis-related genes from [232]GSE134420 and [233]GSE77298,
   revealing a strong correlation between the ENO1 gene and ferroptosis
   [234][71]. Similarly, Zhang et al. utilized the [235]GSE30718 and
   [236]GSE139061 datasets to identify RPS7 as a diagnostic marker for
   ferroptosis, subsequently confirming its expression and pro-apoptotic
   effects [237][72]. Similar findings for these key genes were noted in
   our validation set. Further experimental validation is needed to
   clarify the functional significance of these genes in ferroptosis. This
   exploration might provide a pathway to uncover new molecular mechanisms
   underlying ICM and facilitate the search for prognostic biomarkers.

   In recent years, non-invasive measurements, as an ideal approach
   pursued in clinical diagnostics, offer new possibilities for the early
   detection of ICM through the evaluation of genetic biomarkers. Although
   our research is predominantly based on cardiac tissue data, existing
   literature indicates that certain biomarkers directly associated with
   changes in cardiac conditions can also be reflected in blood [238][73].
   Key genes identified in our study, such as ENO1, have been reported to
   exhibit alterations in blood, suggesting its potential application in
   non-invasive assessments. [239][74]. Although the current research on
   the non-invasive measurement of genes in ischemic heart disease is
   still evolving, preliminary results indicate the potential of gene
   expression levels as non-invasive biomarkers. In the future, as
   technology continues to advance and research deepens, gene-related
   non-invasive measurement methods will play an increasingly important
   role in the early diagnosis and therapeutic monitoring of ischemic
   heart disease.

   To investigate the potential signaling pathways linked to DEGs
   associated with ferroptosis, we performed a functional enrichment
   analysis. Our GO analysis and GSEA indicated that oxidative stress
   might contribute to ferroptosis in ICM. This finding supports
   conclusions from earlier research. Furthermore, the KEGG analysis
   identified particular enriched pathways, notably the TNF signaling
   pathway, which is significantly associated with immune responses in
   ICM. Results from a microarray study on intracerebral hemorrhage using
   the [240]GSE24265 dataset suggest that the TNF signaling pathway is
   crucial in mediating ferroptosis after cerebral hemorrhage, aligning
   with our hypotheses [241][75]. Moreover, recent studies have
   demonstrated that the TNF signaling pathway contributes to the
   protection of fibroblasts against ferroptosis by enhancing cystine
   uptake and stimulating the biosynthesis of GSH, thus facilitating the
   advancement of rheumatoid arthritis [242][76]. However, the exact
   mechanisms by which the TNF signaling pathway influences
   ferroptosis-related cardiomyocyte death in ICM are still unclear.
   Consequently, further prospective research is imperative for a thorough
   exploration of this phenomenon. In conclusion, our findings emphasize
   the connection between ferroptosis and ICM, highlighting the critical
   need for further experimental studies to clarify the underlying
   mechanisms.

   Oxidative stress occurs when there is an imbalance between ROS
   production and the effectiveness of antioxidant systems or free-radical
   scavengers, leading to the activation of various TFs and
   pro-inflammatory genes [243][21]. The underlying causes for the
   overexpression or downregulation of these central genes within the
   myocardium and other tissues remain unclear. To investigate this
   phenomenon, we employed bioinformatics approaches to predict potential
   TFs. Our findings revealed the identification of ten TFs, namely
   proliferation-associated 2G4(PA2G4), protein arginine methyltransferase
   3(PRMT3), zinc finger protein 888(ZNF888), zinc finger protein
   146(ZNF146), NME/NM23 nucleoside diphosphate kinase 2(NME2), JunD
   proto-oncogene, AP-1 transcription factor subunit(JUND), zinc finger
   protein 581(ZNF581), MYC proto-oncogene(MYC), CCAAT enhancer binding
   protein zeta(CEBPZ), and E2F transcription factor 4(E2F4).
   Interestingly, many hub genes were regulated by the same TF, indicating
   that they might have similar biological functions. JUND, recognized as
   a downstream transcription factor of linc00976, was shown to facilitate
   the transcription of linc00976, which is crucial for promoting
   tumorigenesis and metastasis, as well as inhibiting ferroptosis through
   the modulation of the miR-3202/GPX4 axis [244][77]. Zhang et al. also
   identified JUND as a TF associated with ferroptosis through
   bioinformatics analysis. Numerous studies have substantiated that the
   aberrant expression of members of the Myc TF family could impede
   ferroptosis [245][78], [246][79]. For instance, Zhao et al.
   demonstrated that the deprivation of glutamine could stimulate the
   expression of the c-Myc TF, which in turn promotes the transcription of
   Nrf[2], thereby sustaining GSH synthesis and inhibiting ferroptosis
   [247][80]. The other TFs identified in this investigation require
   further validation and exploration. Moreover, our drug sensitivity
   analysis showed that NACA negatively correlates with several
   small-molecule drugs in both datasets. Consequently, NACA might serve
   as a novel biomarker for predicting drug responses in ICM in clinical
   settings and could provide a new perspective for the future management
   of ICM. Cardiac ischemia has been observed in approximately 5 % of
   patients treated with paclitaxel, a microtubule inhibitor widely
   utilized in the chemotherapy of various cancers, including ovarian,
   breast, and colorectal cancers [248][81].Additionally, another study
   indicated that paclitaxel might significantly contribute to enhancing
   cardiac functional recovery during reperfusion [249][82]. Tamoxifen, a
   selective estrogen receptor modulator, has been effectively employed in
   breast cancer treatment [250][83]. One finding suggested that tamoxifen
   might have cardioprotective effects against I/R injury in rat models
   [251][84]. Our data mining from drug databases indicated that elevated
   levels of NACA might increase sensitivity to both paclitaxel and
   tamoxifen. However, further investigation into the functional
   mechanisms is warranted.

5. Limitations

   This study analyzed two microarray datasets to identify potential hub
   genes related to ferroptosis in ICM, but it has certain limitations.
   The limitation to just two datasets was due to current learning
   conditions. Future research will focus on including more datasets to
   address the sample size constraints. Additionally, relying on a limited
   sample from GEO database, without analyses from other databases, has
   restricted the validation of our findings. Increasing the number of
   tissue samples is crucial for further validation. Furthermore, the in
   vitro validation of ferroptosis requires specific experimental
   conditions and advanced technical equipment, which have not yet been
   acquired. Given the constraints of available resources, alternative
   bioinformatics analysis methods were employed in this study, the
   results of which have been partially corroborated by existing
   literature. Once the necessary conditions are met, future studies will
   combine in vivo and in vitro methods to clarify the expression and
   clinical significance of the most important DEGs, especially in
   ischemic regions. Lastly, due to the intrinsic limitations associated
   with bioinformatics analyses in predicting drug susceptibility, coupled
   with the absence of clinical data, prospective validation studies are
   essential to substantiate our predictions regarding drug sensitivity.

6. Conclusion

   In conclusion, our findings provide valuable insights into the DEGs
   related to ferroptosis in ICM through bioinformatics analysis([252]Fig.
   9). Genes associated with ferroptosis might offer a promising avenue
   for future ICM research. We have identified 11 key genes (AHCY, CCT3,
   CCT4, CCT5, CCT6A, EIF4A1, ENO1, NACA, RPLP0, RPS2, and RPS7) that
   might serve as biomarkers or therapeutic targets for ICM. Further
   research in ischemic tissues is crucial to clarify the functional roles
   of these core genes. This will enhance our understanding of their
   implications in ferroptosis and oxidative stress. Furthermore, using
   various experimental methods will help identify new therapeutic targets
   and biomarkers, contributing to innovative treatment strategies for
   ICM.

Fig. 9.

   [253]Fig. 9
   [254]Open in a new tab

   A schematic illustration depicting the relationship between ferroptosis
   and ICM. This illustration shows how ferroptosis and oxidative stress
   interact in the development of ICM. Ferroptosis occurs due to factors
   like iron overload and hypoxia, and is characterized by lipid
   peroxidation. Key genes such as AHCY, CCT4, CCT5, CCT6A, ENO1, NACA,
   RPLP0, and RPS7 play crucial roles in processes that may worsen ICM.
   Elevated levels of ROS promote ferroptosis and lead to
   pathophysiological changes in ICM.These changes can result in
   complications like myocardial ischemia and heart failure. The TNF
   signaling pathway is emphasized for its crucial role in connecting
   ferroptosis and oxidative stress to the development of ICM.

CRediT authorship contribution statement

   Huilin Liu: Writing – review & editing, Writing – original draft, Data
   curation. Yuan Xu: Investigation, Formal analysis. Yuanmei Liu: Writing
   – review & editing, Investigation, Data curation. XueJun Han: Writing –
   review & editing, Writing – original draft. Liping Zhao: Writing –
   review & editing, Writing – original draft. Yixuan Liu: Visualization,
   Software. Fuchun Zhang: Funding acquisition, Conceptualization. Yicheng
   Fu: Validation, Methodology, Investigation, Formal analysis,
   Conceptualization.

Declaration of competing interest

   The authors declare that they have no known competing financial
   interests or personal relationships that could have appeared to
   influence the work reported in this paper.

Funding

   This work was supported by grants from the Key Clinical Program of
   Peking University Third Hospital (Grant No. BYSYDL2021022 and Grant
   No.BYSYDL202300106).

Footnotes

   ^Appendix A

   Supplementary data to this article can be found online at
   [255]https://doi.org/10.1016/j.ijcha.2024.101584. [256]Supplementary
   Fig. S1 shows gene expression distributions within all samples and
   batch effect correction. [257]Supplementary Fig. S2 shows the
   distribution differences of 53 ferroptosis-related differential genes
   in [258]GSE5406. [259]Supplementary Fig. S3 shows the distribution
   differences of 53 ferroptosis-related differential genes in
   [260]GSE57338. [261]Supplementary Fig. S4 shows pathway mapping of
   ferroptosis-related differential genes. (a) Network diagram of the GO
   enrichment analysis. (b) Network diagram of the KEGG enrichment
   analysis. Pathway names are marked with yellow nodes and connected to
   the genes involved (grey nodes) with different colour-coded lines. (c)
   Pathway enrichment heat map of GO enrichment analysis. (d) Pathway
   enrichment heat map of KEGG enrichment analysis. The vertical axis
   corresponds to the enriched pathway, and the horizontal axis
   corresponds to the gene involved. The black colour represents the
   enrichment result as “yes”. (e) Network diagram of GO enrichment
   analysis. (f) Network diagram of KEGG enrichment analysis.
   Significantly enriched pathways are presented as nodes, and
   interrelationships are represented by grey lines. [262]Supplementary
   Fig. S5 shows protein-Protein network analysis. (a) The PPI network of
   the 53 ferroptosis-related differentially expressed genes. (b)
   Interaction network diagram of the 11 hub genes. (c) A subnetwork of
   hub genes extracted from the PPI network. [263]Supplementary Fig. S6
   shows TF–hub-gene prediction scores. The x-axis represents the
   TF-score, the y-axis represents the name of TFs, and the color
   represents the number of regulated target genes. TF, transcription
   factor. [264]Supplementary Fig. S7 shows hub Interaction network of hub
   genes and TFs. (a) TF-hub-gene interaction network diagram. Grey
   represents the hub gene, and pink represents the TFs. (b) TF-target
   gene interaction network diagram, where yellow represents the TFs, grey
   represents target genes, and the right shows the TFs with interaction.
   (c) Schematic of the genes targeted by the TFs. Target genes are grey
   and TFs are yellow. [265]Supplementary Fig. S8 shows drug sensitivity
   analysis. (a) Bubble diagram of drug sensitivity analysis in the CCLE
   database. (b) Bubble diagram of the drug sensitivity analysis of the
   GDSC database. [266]Supplementary Table S1 shows rank test results of
   53 ferroptosis-related differential genes in [267]GSE5406 and
   [268]GSE57338, respectively. [269]Supplement Table S2 shows GO analysis
   results and significant GO terms of the modified ferroptosis-related
   DEGs in ICM. [270]Supplement Table S3 shows results of GSEA in
   [271]GSE5406 and [272]GSE57338.

Appendix A. Supplementary data

   The following are the Supplementary data to this article:
   Supplementary Data 1
   [273]mmc1.docx^ (12.2KB, docx)
   Supplementary Data 2
   [274]mmc2.pdf^ (3.3MB, pdf)
   Supplementary Data 3
   [275]mmc3.docx^ (49.9KB, docx)

References