Abstract

   Acute myocardial infarction (AMI) is a major contributor to
   cardiovascular-related mortality, and early diagnosis is crucial for
   effective treatment and better outcomes. While several biomarkers have
   been explored for AMI, there remains a need for reliable, non-invasive
   biomarkers that can accurately differentiate AMI patients from healthy
   individuals. This study aims to identify potential mRNA biomarkers in
   peripheral blood that could aid in the diagnosis and monitoring of AMI.
   We performed transcriptomic analysis of blood samples from 81
   individuals, including 16 healthy controls, 58 AMI patients, and 7
   post-treated AMI individuals. Through a combination of Sparse Partial
   Least Squares-Discriminant Analysis (sPLS-DA), random forest (RF),
   Weighted Gene Co-expression Network Analysis (WGCNA), and LASSO
   regression, we identified mRNA markers that are significantly
   correlated with AMI. Specifically, the mRNA expressions of ANKRD52,
   ART1, NRP2, and PPP1R15A were elevated in AMI patients, whereas
   BAIAP2L1 and CCNE1 were downregulated. However, while these mRNA
   biomarkers show potential for distinguishing AMI patients from healthy
   individuals, further studies are needed to confirm their clinical
   applicability.

Supplementary Information

   The online version contains supplementary material available at
   10.1038/s41598-025-92757-4.

   Keywords: Acute myocardial infarction (AMI), mRNA biomarkers, Machine
   learning, Diagnostic screening

   Subject terms: High-throughput screening, Diagnostic markers,
   Cardiovascular biology

Introduction

   Acute myocardial infarction (AMI) represents a significant global
   health challenge, with rising incidence and mortality rates,
   particularly in regions with low to middle socioeconomic indices^[38]1.
   Early detection of AMI is imperative for the timely initiation of
   therapeutic interventions to reduce mortality rates. Although troponin
   and creatine kinase-MB are the conventional markers for AMI diagnosis,
   their specificity is limited, leading to potential false positives in
   non-cardiac conditions^[39]2. Consequently, the scientific community
   continues to explore new biomarkers that offer higher sensitivity and
   specificity for the early detection of AMI.

   Previous investigations have explored AMI-specific diagnostic markers
   through mRNA and single-cell transcriptomics; however, the progression
   of AMI and cardiac repair is a protracted and dynamic biological
   process. This complex process spans from the initial transformation of
   healthy cardiomyocytes under hypoxic conditions, which alters energy
   metabolism and induces oxidative stress, to the eventual myocardial
   injury repair. Biomarker levels notably fluctuate throughout these
   stages, reflecting the dynamic nature of AMI pathophysiology. A study
   monitoring AMI patients over a 20-week post-event observation period
   revealed that levels of cardiac-specific miRNA-1 and
   endothelial-specific miRNA-126 in circulation were significantly
   elevated compared to healthy control, with a subsequent decline during
   the observation period^[40]3. In ischemic cardiac tissue, HIF-1α
   protein accumulates persistently, and vascular endothelial growth
   factor A (VEGFA) is induced in the acute phase of myocardial infarction
   but not during the chronic phase^[41]4. There is an upregulation of
   insulin-like growth factor 2 (IGF2) and angiopoietin 2 (ANGPT2) in the
   infarcted region, while expressions of VEGFA and fms-related tyrosine
   kinase 1 (FLT1, also known as VEGF receptor 1) are diminished^[42]5.
   Single-cell sequencing of peripheral blood has revealed an upregulation
   of the IL-1 signaling pathway following ST-Elevation Myocardial
   Infarction (STEMI), indicating an ongoing heart failure process^[43]6.
   Analysis of two GEO high-throughput datasets reveals that
   post-myocardial infarction, hsa-miR-330-3p exerts regulatory effects on
   MMP2, leading to macrophage infiltration into atherosclerotic lesions
   and promoting the secretion of numerous factors such as MMP-2. These
   factors contribute to endothelial cell damage by cleaving extracellular
   matrix components, collagen, and elastin, thereby accelerating the
   formation of intravascular thrombi^[44]7. These findings suggest that
   the potential origin of variability in AMI biomarker studies is
   attributed to the differing gene expression patterns in myocardial
   tissues and peripheral blood^[45]7. During myocardial infarction,
   mononuclear cells, macrophages, lymphocytes, and smooth muscle cells
   accumulates and release of various cytokines and inflammatory
   biomarkers^[46]8. Single-cell transcriptomics provides the necessary
   resolution to distinguish cell-specific responses, which are often
   masked in analyses of bulk tissue^[47]9. The complexity of myocardial
   infarction pathophysiology, the dynamic nature of gene expression
   patterns, and the heterogeneity of patient populations can contribute
   to the observed variability of identified biomarkers. As myocardial
   infarction progresses, transcriptional activity undergoes selective
   regulation, necessitating the identification of more relevant markers
   through the assessment of myocardial recovery at various post-treatment
   stages.

   The advent of machine learning has markedly altered the stratification
   of risk and the prediction of mortality in AMI. Chang et al.^[48]10
   identified a five-feature model including troponin I, HDL cholesterol,
   HbA1c, anion gap, and albumin as potential biomarkers for the early
   detection and treatment of AMI. This model utilizes feature selection
   combined with machine learning techniques, building upon established
   risk factors for AMI and cardiovascular diseases. A study has
   corroborated the expression specificity of immune-related genes in AMI,
   which can facilitate diagnosis in clinical settings through the
   development of various machine model^[49]11. Moreover, the application
   of machine learning extends beyond biomarkers. Employing algorithms
   that integrate multimodal data, including electrocardiographic
   waveforms, demographic information like gender and age, have been shown
   to enhance the diagnostic accuracy for myocardial ischemia or
   infarction^[50]12. Despite these advances and promising results,
   characterized by high performance on receiver operating curves, these
   innovative models necessitate further clinical trials to affirm their
   applicability and efficacy in real-world medical settings.

   In this study, we performed a transcriptomic analysis on peripheral
   blood samples from AMI patients, individuals seven days post-treatment,
   and healthy controls. This analysis unveiled AMI-associated
   transcriptomic signatures, including specific mRNA expression patterns
   linked to key signaling pathways. Employing a combination of data
   analytical techniques and machine learning, we executed a multi-step
   screening process aimed at identifying a set of potential mRNA
   biomarkers correlated with AMI.

Materials and methods

Study design and sample collection

   This study was conducted in accordance with the ethical standards and
   approval of the Kunming University of Science and Technology (KUST) and
   the People’s Hospital of Lijiang. All procedures involving human
   participants were in compliance with the ethical standards of the
   research committee of KUST and People’s Hospital of Lijiang. This study
   was conducted in accordance with the principles of the Declaration of
   Helsinki and informed consent was obtained from all participants
   involved in this research. For those participants who were unable to
   provide consent due to their medical condition, consent was obtained
   from their legal guardians. All subjects were informed about the nature
   of the study, the procedures involved, potential risks, and their right
   to withdraw at any time without affecting their medical care.

   Patients presenting with AMI were recruited and treated at the People’s
   Hospital of Lijiang City, Yunnan Province. The inclusion criteria for
   participation in this study were: (1) a clinical diagnosis of AMI,
   meeting the standards defined in the “Guidelines for the Diagnosis and
   Treatment of AMI”; (2) long-term residents living in Lijiang for more
   than six months; (3) both male and female patients aged 18 years or
   older; (4) no known hematological diseases or severe liver and kidney
   dysfunction; (5) no family history of hereditary tumors; (6) no history
   of radiation therapy or chemotherapy; (7) no occupational exposure to
   radioactive materials, toxic gases, or other carcinogens; and (8)
   written informed consent signed by the patient or their legal
   representative.

   Patients were excluded from the study if they met any of the following
   criteria: (1) a history of diagnosed hematological diseases or severe
   renal or hepatic dysfunction; (2) recent (within the past six months)
   coronary intervention treatment (such as stent placement or coronary
   artery bypass surgery); (3) currently participating in other
   interventional clinical studies involving AMI; (4) presence of other
   severe comorbidities, such as active severe infections, advanced
   malignancies, or immunodeficiency diseases; (5) women who are pregnant
   or breastfeeding; and (6) any other conditions deemed unsuitable for
   participation by the researchers, such as poor compliance or
   significant confounding factors.

   Upon confirmation of eligibility, detailed demographic information,
   chronic disease history, laboratory test results, and diagnostic data
   were recorded for each patient, and peripheral blood samples were
   obtained at the time of admission. Peripheral blood was also collected
   from a subset of seven patients 7–10 days into treatment when their
   condition had stabilized. The healthy control group was publicly
   recruited from the local population by our department, consisting of
   individuals aged 18 or older, excluding those with underlying
   cardiovascular conditions, pregnant individuals, and patients with
   cancer. After drawing peripheral blood from participants, each sample
   was immediately mixed with threefold volume of Trizol (Qiagen, Germany)
   and stored at -80 °C for subsequent use.

RNA-seq and data analysis

   Genomic RNA from all enrolled samples of peripheral blood was extracted
   using the TRNzol Universal Reagent kit (Tiangen, China). The extracted
   RNA was quantified and its quality assessed through: (1) Measuring
   sample concentration with the Qubit 4.0 Fluorometer (Invitrogen, USA)
   and the Equalbit RNA BR Assay Kit (Vazyme, China). (2) Evaluating RNA
   integrity using the Qsep1 Bio-fragment Analyzer (Guangding, China).
   High-quality RNA with a concentration ≥ 80 ng/µl, and a total
   yield ≥ 1 µg was then used for subsequent library construction. A total
   of 3 µg of RNA was employed for the library construction process, which
   included mRNA enrichment with Oligo(dT) magnetic beads (Yesen, China),
   fragmentation, synthesis of first- and second-strand cDNA, purification
   with AMPure XP beads (Beckman, USA), end-repair, A-tailing, adaptor
   ligation, and PCR enrichment. Completed libraries were quantified and
   quality-assessed using Qubit 4.0, the Agilent 2100 Bioanalyzer (Agilent
   Technologies, USA), and the Bio-RAD CFX 96 Real-Time System (Biorad,
   USA), ensuring library concentrations exceeded 10 nM for subsequent
   sequencing. Sequencing was performed on the NovaSeq 6000 S4 platform
   (Illumina, USA) using the Illumina NovaSeq 6000 S4 Reagent Kit V1.5,
   generating PE150 reads. Quality control of all sequencing data was
   conducted using FastQC software for analysis and filtration. Human
   sequences were removed by mapping to the human reference genome
   (GRCh38.p13 [51]https://www.ncbi.nlm.nih.gov/assembly/2334371) using
   BWA (BurrowsWheeler alignment).

   In this study, we conducted various analyses, including RNA-seq
   analysis, sPLS-DA analysis, WGCNA analysis, and animal analysis. For
   the detailed methodologies of these analyses, please refer to the
   Supplementary Methods.

Real‑time quantitative polymerase chain reaction (RT-qPCR)

   Total RNA was isolated using TRIzol (Invitrogen, USA), and 1 µg RNA was
   used for reverse transcription using a PrimeScript RT Reagent Kit
   (TaKaRa, JPN) according to the manufacturer’s instructions. RT–qPCR was
   performed by using TB Green Premix Ex Taq II (TaKaRa, JPN). The samples
   were processed using a Thermo Q6 Real-Time System. Five genes were
   selected for validation as potential biomarkers for AMI. Three
   replicates were run for each sample. The primer sequences are shown in
   Table.S1.

Statistical analysis

   Statistical analyses were performed using R (Version 4.2). Quantitative
   data were analyzed as appropriate based on their distribution, and
   categorical variables were assessed accordingly. A p value of ≤ 0.05
   was considered statistically significant.

   For detailed methodologies related to statistical tests, including the
   treatment of normally and non-normally distributed data, as well as the
   handling of categorical variables, please refer to the Supplementary
   Methods ‘Statistical Analysis’ part.

Results

Subject characteristics

   Table [52]1 presents the demographic and clinical profiles of the
   study’s participants, comparing healthy controls, AMI patients, and
   those receiving AMI treatment. The study workflow is illustrated in
   Fig. [53]1. A total of 87 participants were recruited, and clinical
   data - including age, sex, comorbidities, and clinical AMI indicators
   were collected. Peripheral blood samples were obtained from all
   subjects, followed by transcriptomic sequencing. To ensure data
   accuracy, individuals with incomplete clinical information (4 from the
   control group) and those with poor sequencing quality (2 from the AMI
   group) were excluded. Ultimately, 58 confirmed AMI patients, 7
   post-treatment AMI patients, and 16 healthy controls were included in
   the analysis. After sequencing, a series of analytical steps were
   performed, including differential expression analysis, WGCNA, sPLS-DA,
   and random forest analysis, culminating in LASSO regression for feature
   selection and validation in a mouse model. This integrated approach
   aimed to identify key gene expression differences across groups to
   uncover potential biomarkers and therapeutic targets for AMI. Cardiac
   biomarkers including Ctn1, CKMB, and MYO were measured between 7 and 10
   days post-treatment, while other data were collected at hospital
   admission. The AMI group (56.21 ± 12.20) was older than the control
   group (33.94 ± 5.69) (p < 0.001). Given the multi-ethnic composition of
   the Lijiang region in Yunnan, the proportions of minority groups were
   catalogued within different cohorts, with the majority being Han,
   followed by Bai, Naxi, Tibetan, Yi and others ethnicities. Hypertension
   prevalence was higher in the AMI group (41.38%) compared to the control
   group (0.00%), while the prevalence of diabetes showed no significant
   difference across groups (P = 0.136). Smoking history was significantly
   more common in the AMI group (60.34%), compared to an absence of
   smoking history in the control group (0.00%) (P < 0.001). The primary
   diagnosis within the AMI group was Inferior STEMI (55.17%). After
   treatment, the levels of Ctn1, CKMB, and MYO significantly
   decreased(P < 0.001), suggesting therapeutic efficacy.

Table 1.

   Clinical characteristics of participant.
   Variables AMI (n = 58) Control (n = 16) Treatment (n = 7) P
   Age, Mean ± SD 56.21 ± 12.20 33.94 ± 5.69 57.14 ± 16.57 < 0.001
   Sex, n (%) < 0.001
    Female 10 (17.24) 14 (87.50) 2 (28.57)
    Male 48 (82.76) 2 (12.50) 5 (71.43)
   Minority, n (%) 0.075
    Bai 12 (20.69) 0 (0.00) 0 (0.00)
    Han Chinese 29 (50.00) 5 (31.25) 4 (57.14)
    Naxi 8 (13.79) 8 (50.00) 0 (0.00)
    Tibetan 5 (8.62) 0 (0.00) 0 (0.00)
    Yi 0 (0.00) 0 (0.00) 0 (0.00)
    Other ethnic groups 4 (6.90) 3 (18.75) 3 (42.86)
   Hypertension, n (%) 0.002
    No 34 (58.62) 16 (100.00) 4 (57.14)
    Yes 24 (41.38) 0 (0.00) 3 (42.86)
   Diabetes, n (%) 0.136
    No 47 (81.03) 16 (100.00) 7 (100.00)
    Yes 11 (18.97) 0 (0.00) 0 (0.00)
   Smoking, n (%) < 0.001
    No 23 (39.66) 16 (100.00) 3 (42.86)
    Yes 35 (60.34) 0 (0.00) 4 (57.14)
   Alcohol drinking, n (%) 0.219
    No 43 (74.14) 16 (100) 5 (71.43)
    Yes 15 (25.86) 0 (0.00) 2 (28.57)
   STEMI type, n (%) < 0.001
    Anterior STEMI 14 (24.14) NA 5 (71.43)
    Extensive anterior STEMI 11 (18.97) NA 0 (0.00)
    Inferior STEMI 32 (55.17) NA 2 (28.57)
    None 0 (0.00) 16 (100.00) 0 (0.00)
    Others 1 (1.72) 0 (0.00) 0 (0.00)
   Chest pain duration (Min), M (Q₁, Q₃) 360.00 (144.00, 592.75) NA 250.00
   (123.00, 757.00) < 0.001
   Ctn1 (Ng/Ml), M (Q₁, Q₃) 5.79 (0.83, 48.23) NA 5.00 (3.15, 11.99)
   < 0.001
   CKMB (U/L), M (Q₁, Q₃) 84.00 (30.00, 291.00) NA 22.00 (19.50, 34.00)
   < 0.001
   MYO (ug/L), M (Q₁, Q₃) 317.30 (71.88, 991.17) NA 41.30 (36.05, 46.30)
   < 0.001
   TC, M (Q₁, Q₃) 5.11 (4.24, 6.23) NA NA –
   TG, M (Q₁, Q₃) 2.08 (1.16, 2.78) NA NA –
   ApoA, M (Q₁, Q₃) 1.21 (1.06, 1.35) NA NA –
   LDL-C, M (Q₁, Q₃) 2.99 (2.48, 3.85) NA NA –
   Creatinine, M (Q₁, Q₃) 77.46 (65.00, 89.12) NA NA –
   Urea Nitrogen, M (Q₁, Q₃) 5.15 (4.50, 6.57) NA NA –
   SaO2, M (Q₁, Q₃) 94.00 (93.00, 95.00) NA NA –
   RCA, M (Q₁, Q₃) 0.50 (0.00, 0.99) NA NA –
   LAD, M (Q₁, Q₃) 0.80 (0.43, 0.98) NA NA –
   LCX, M (Q₁, Q₃) 0.40 (0.00, 0.70) NA NA –
   LAX, M (Q₁, Q₃) 0.00 (0.00, 0.00) NA NA –
   [54]Open in a new tab

   SD: standard deviation; M: Median; Q₁: 1st Quartile; Q₃: 3st Quartile;
   STEMI: ST-Elevation Myocardial Infarction.

   Ctn1(Ng/mL): Cardiac Troponin I; CKMB (U/L): Creatine Kinase-MB; MYO
   (µg/L): Myoglobin; TC: Total Cholesterol; TG: Triglycerides; ApoA:
   Apolipoprotein A; LDL-C: Low-Density Lipoprotein Cholesterol; SaO2:
   Arterial Oxygen Saturation.

   RCA, LAD, LCX, and LAX are the stenosis of the contrast findings, and
   the values are in percentage, respectively. RCA: Right Coronary Artery;
   LAD: Left Anterior Descending artery; LCX: Left Circumflex artery; LAX:
   Left coronary Artery axis.

Fig. 1.

   [55]Fig. 1
   [56]Open in a new tab

   Presented the schematic diagram of the study.

Pairwise differential expression analysis reveals distinct mRNA signatures
and treatment effects in AMI

   In our RNA-seq data analysis, we compared gene expression across three
   groups of participants: (1) differences in gene expression between AMI
   patients and healthy controls, (2) gene expression in patients
   post-treatment compared to AMI patients, and (3) comparisons of gene
   expression between post-treatment patients and healthy controls.
   Initially, Principal component analysis (PCA) was employed to
   investigate mRNA profiles among individual samples and groups. This
   unsupervised multivariate approach (Fig. [57]2A) generates principal
   component axes capable of elucidating the variability in the data
   without prior knowledge of the sample groups. Possibly due to sample
   heterogeneity, the three groups (control, AMI, and post-treatment)
   could not be distinctly segregated based on their overall
   transcriptomic landscape. To delve deeper into the data and uncover
   mRNAs associated with AMI, we subsequently engaged in a supervised
   analysis, sPLS-DA, a method that identifies the most discriminative
   mRNAs within the dataset. Through this analysis, we distinguished the
   groups primarily along component 1, as illustrated in the dot plot of
   component 1 versus component 2 (Fig. [58]2B). Although, there was no
   clear separation into molecular subtypes, our analysis revealed
   significant differences in gene expression: 1067 genes between AMI
   patients and controls (Fig. [59]2C), 477 genes between post-treatment
   and AMI individuals (Fig. [60]2D), and 567 genes between post-treatment
   individuals and controls (Fig. [61]2E). The threshold for differential
   gene screening among groups was set as an absolute log2 fold change
   greater than 2 and p-value < 0.05. The differentially expressed genes
   (DEGs) identified in pairwise comparisons among the three groups are
   detailed in Table.S2 (Original). Certain mRNAs were consistently
   modulated across both comparative analyses (Fig.[62]S1). Certain mRNAs,
   including HBEGF, G0S2, EREG, FOSB, PLK2, NR4A2, HEY1, SFN, PER1,
   RNU6-415P, LINC00664, CSAD and CNTNAP3C, were significantly upregulated
   in the AMI group compared to the controls. However, the mRNAs CSAD,
   IL32, RPL36, VAMP5, ZNF48, RPS26, UQCC3, RPL13P12, and ZNF593OS showed
   no significant change compared to the AMI group after treatment
   (Table.S3). Taking into account the significant age differences between
   the AMI group and the control group, we included age as a covariate and
   conducted pairwise differential comparison analyses among the three
   groups. A total of 3,033 genes were identified between AMI patients and
   controls (Fig. [63]S2A), 941 genes between post-treatment individuals
   and AMI patients (Fig. [64]S2B), and 1,342 genes between post-treatment
   individuals and controls (Fig. [65]S2C). Several differentially
   expressed genes exhibited consistent expression patterns regardless of
   age adjustment; for instance, FOSB, G0S2, and CNTNAP3C were
   significantly upregulated in the AMI group compared to controls and
   remained significantly upregulated upon re-analysis, with a notable
   downregulation observed in the treatment group compared to AMI
   patients. However, another set of genes, such as UTY, ANKRD36BP1, and
   KDM5D, which were significantly upregulated in the AMI group relative
   to controls, displayed altered expression patterns when age was
   considered as a covariate [Table [66]S2 (Age as a covariate)].In the
   analysis accounting for age as a covariate, we identified a new set of
   significantly downregulated genes, including PROCR (log2FoldChange =
   -2.36, p = 0.0488), LINC02987 (log2FoldChange = -2.36, p = 0.0215), and
   MAD2L2 (log2FoldChange = -2.52, p = 0.0012). Notably, MED18 and ASIC1
   were also significantly downregulated in both analyses.

Fig. 2.

   [67]Fig. 2
   [68]Open in a new tab

   Transcriptomic analysis was conducted on peripheral blood samples from
   AMI and treatment groups. (A) Depicted the principal component analysis
   (PCA) plot based on transcriptomics data. (B) The sPLS-DA demonstrated
   moderate clustering between the three groups. (C) The volcano plots
   displayed differentially expressed genes (DEGs) between the AMI group,
   the post-treatment group and the healthy control group. The black line
   represented the cutoff line with indicated significance criteria.
   Points with absolute log fold-change ≥ 2 and P < 0.05 were shown in
   blue, points with absolute log fold-change ≤ -2 and P ≤ 0.05 were in
   red, and the rest were in green: (C) AMI versus healthy controls, (D)
   Post-treatment of AMI versus AMI, and (E) Post-treatment of AMI versus
   healthy controls. AMI Acute Myocardial Infarction, FC fold change,
   Treat Post-treatment of AMI.

Deciphering molecular pathways and gene functions through enrichment analysis

   To elucidate the roles of identified genes, we conducted enrichment
   analyses using Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene
   Ontology (GO) enrichment analyses. GO analysis revealed that, compared
   to the control group, the upregulated DEG in the AMI group were
   enriched in 795 terms across Biological Process (BP), Cellular
   Component (CC), and Molecular Function (MF) categories (p < 0.05). The
   top 10 enriched terms are depicted in Fig. [69]3A. Notably, BP was
   predominantly enriched in positive regulation of calcium ion transport,
   CC showed enrichment in platelet alpha granule lumen, and MF was
   enriched for platelet-derived growth factor binding. In the comparison
   between post-treatment and AMI, upregulated DEGs were associated with
   an enrichment of 159 terms (p < 0.05), which included cell junction
   disassembly in BP, blood microparticle in CC and antigen binding in MF.
   Downregulated DEGs in this comparison were associated with 721
   significant terms (p < 0.05), including regulation of epithelial cell
   proliferation in BB, semaphorin receptor complex in CC, and in
   transmembrane receptor protein tyrosine kinase activity in MF
   (Fig. [70]3B). When comparing treated individuals to controls, the
   upregulated genes were related to 284 significant terms (p < 0.05),
   encompassing regulation of epithelial cell proliferation in BP,
   platelet alpha granule in CC, and in type II transforming growth factor
   beta receptor binding in MF (Fig. [71]3C).

Fig. 3.

   [72]Fig. 3
   [73]Open in a new tab

   GO functional enrichment and pathway enrichment analyses were conducted
   on DEGs. (A–C) The top 5 GO terms associated with the Molecular
   Functions (MF), Biological Processes (BP), and Cellular Compartments
   (CC) in both upregulated and downregulated DEGs were identified. (A) In
   the comparison between AMI and healthy controls, (B) Post-treatment of
   AMI versus AMI, and (C) Post-treatment of AMI versus healthy controls.
   (D–F) KEGG enrichment analyses were performed on the upregulated and
   downregulated DEGs. The top 10 were shown here. (D) For the comparison
   between AMI and healthy controls, (E) Post-treatment versus AMI, and
   (F) Post-treatment of AMI versus healthy controls.

   In the pathway enrichment analysis facilitated by KEGG, DEGs between
   AMI and control groups highlighted enrichment in the top 10 pathways
   (Fig. [74]3D). The upregulated DEGs were predominantly enriched in
   pathways associated with Amoebiasis, Nitrogen metabolism and Glycine,
   serine, and threonine metabolism. Notably, genes such as CXCL8
   (log2FC = 3.38, p. adjust < 0.01) and FOSB (log2FC = 5.86, p < 0.01,
   p.adjust < 0.01), which were significantly upregulated in AMI group,
   partook in essential immune-related pathways (Table.S4). Downregulated
   DEGs were enriched in Cholinergic synapse, Relaxin signaling pathway
   and Neuroactive ligand-receptor interaction. Comparative pathway
   enrichment of DEGs between Treatment and AMI groups were depicted in
   Fig. [75]3E, with upregulated genes enriched in Inositol phosphate
   metabolism, Leishmaniasis and Hypertrophic cardiomyopathy, while
   downregulated genes were enriched in IL-17 signaling pathway,
   Biosynthesis of amino acids and Leishmaniasis. The top 10 pathways of
   DEGs enrichment between Treatment and Control comparisons are presented
   in Fig. [76]3F, where upregulated differential genes were enriched in
   TGF-beta signaling pathway, Osteoclast differentiation and non-small
   cell lung cancer. Downregulated DEGs were enriched in Ribosome,
   Coronavirus disease COVID-19 and African trypanosomiasis.

   Incorporating age as a covariate significantly altered the gene
   enrichment analysis results for AMI patients. Unlike the age-excluded
   analysis, both AMI and control groups showed enrichment in the
   ECM-receptor interaction pathway (Fig. [77]S2C). GO analysis also
   revealed enrichment in the platelet-related pathway, specifically in
   the platelet alpha granule lumen (Fig. [78]S2B), highlighting a notable
   increase in platelet-derived growth factor binding. Additionally, while
   the positive regulation of calcium ion transport pathway was previously
   significant, it lost significance under age consideration, shifting to
   regulation of calcium ion transport. The downregulated genes in the AMI
   group showed significant enrichment in the Relaxin Signaling Pathway,
   with RLN2 being significantly downregulated in both analyses. When
   comparing the treatment group to the AMI group, upregulated genes were
   enriched in fatty acid metabolism. However, in the age-adjusted
   treatment group, several immune-related pathways, such as Primary
   immunodeficiency and Autoimmune thyroid disease, were significantly
   upregulated—findings not observed in the conventional analysis.

   Both analyses revealed enrichment in arginine and proline metabolism
   for downregulated genes in the treatment group compared to the AMI
   group. The age-excluded analysis showed downregulation in the IL-17
   Signaling Pathway and Cytokine-cytokine receptor interaction, which
   were upregulated in both AMI and control groups. However, after
   considering age as a covariate, the enrichment of the Cytokine-cytokine
   receptor interaction pathway decreased.

Sparse partial least squares regression discriminant analysis (sPLS-DA)

   Component 1, encapsulating 38 mRNAs effectively differentiates AMI
   patients from control subjects. Figure [79]4A showed the top 25
   contributors to this component. Similarly, Component 2, comprising 50
   mRNAs detailed in Fig. [80]4B, distinguishes treatment from controls,
   emphasizing the leading 25 mRNAs. The efficacy of these two components
   in classification is underscored by their comprehensive coverage. Data
   revealed that a very good classification was obtained with two
   components. Analysis of receiver operating characteristic (ROC) curves
   conducted for Component 1 revealed that the area under the curve (AUC)
   for discerning AMI from the other groups was 0.9195 (P < 0.01). The
   contribution of individual mRNAs to Component 1 was ordered by their
   importance in mRNA expression, with notable contributors including
   CC3orf52, CTNNB1, REC114, LINC01736, and POLR3E. The ROC curve for
   Component 2 demonstrated an AUC of 0.9974 (P < 0.01) for
   differentiating the AMI group from other groups, and an AUC of 0.9769
   (P < 0.01) for the distinction of the control group from others. The
   top five contributors to Component 2 were ZDHHC3, ZNGB1, ZDHHC6,
   SPATA7, and SLC31A1. It was noteworthy that certain mRNAs, such as
   ZFP36 (log2FC = 1.76, p.adjust = 0.012), were significantly upregulated
   in AMI compared to control and also contributed to Component 2. The
   standalone contribution of the transcriptomic data to the final sPLS-DA
   model was demonstrated by score plots of the first two components,
   indicating the optimal separation capacity of the transcriptomic data
   (Fig. [81]4C).

Fig. 4.

   [82]Fig. 4
   [83]Open in a new tab

   Sparse partial least-squares discriminant analysis. (A, B) Selected
   features were shown in a pyramid bar plot. The loading plot represented
   the top 25 mRNAs contributing to group separation (Left). The area
   under the ROC curve (AUC) values, compared to the grouping level, were
   shown for the three-group classification (Right). (A) The bar plot
   showed the contribution value of each mRNA. (B) sPLS-DA contributions
   to component 2 were depicted. The color in the bar plot represented the
   highest mRNA expression level in the corresponding group. (C) Displayed
   the background prediction plot for the sample’s prediction results.

Identification of the most significant module by WGCNA

   To identify gene modules of biological significance associated with
   AMI, we employed WGCNA to construct a gene correlation network. Outlier
   samples were examined using hierarchical clustering methods, resulting
   in the exclusion of four outliers (Fig. [84]5A, h = 15000). To ensure a
   scale-free co-expression network (Fig. [85]5B), we selected a
   soft-thresholding parameter β = 14, achieving a scale-free fit index
   (R^2) of 0.72. Following the identification and merger of highly
   similar modules, 34 modules were delineated and color-coded
   (Fig. [86]5C). Based on the correlation analysis between modules and
   traits (Fig. [87]5D), the tan module, comprising 207 mRNAs, was
   identified as positively correlated with the clinical trait of chest
   pain duration (r = 0.73, P < 0.01). The plots of module membership and
   gene significance further demonstrated significant correlations within
   the tan module (Fig. [88]5E).

Fig. 5.

   [89]Fig. 5
   [90]Open in a new tab

   The results of weighted gene co-expression network analysis (WGCNA).
   (A) Clustering dendrogram of 81 samples. (B) Analysis of the scale-free
   index for various soft-threshold powers (left). Analysis of the mean
   connectivity for various soft-threshold powers(right). (C) WGCNA was
   performed to identify 28 modules by unsupervised clustering. (D)
   Heatmap of the correlation between the module eigengenes and clinical
   traits of AMI, healthy control, and Post-treatment of AMI. The tan
   module was identified as the positively correlated module with chest
   pain duration (r = 0.73, P < 0.01). (E) The gene significance and
   module membership of the genes in the tan module exhibited a positive
   correlation. (F) KEGG enrichment analyses in tan module.

   To elucidate the functional roles of the genes within the tan module,
   we conducted a KEGG enrichment analysis. The analysis revealed that
   five mRNAs within the tan module—CD44, FLT3LG, CSF3R, CD36, and
   HLA-DPB1—were enriched in the Hematopoietic cell lineage pathway
   (Fig. [91]5F).

Identification of candidate mRNA biomarkers

   Employing a Random Forest (RF) importance score, we prioritized the top
   200 mRNAs, illustrating their contribution to classification
   performance based on disease status (Fig.S3). The top 5 genes with
   significant contributions under this algorithm are POLR3E, ZNF831,
   TRIM17, WSB1, and PABPC1(Fig.S3). The out-of-bag (OOB) estimate of
   error rate was 26.15%, and stability in the ensemble of decision trees
   was reached at approximately 400 trees (Fig.S4). Subsequently, we
   integrated the intersections of the top 200 mRNAs from the RF model,
   the sPLS-DA components 1 and 2, the DEGs of AMI vs. Control, Treatment
   vs. Control, Treatment vs. AMI, and the mRNAs from the tan module
   identified in the WGCNA. The differential genes were derived from the
   analysis that did not consider age as a covariate. The UpSet plot
   revealed an intersection of 92 mRNAs (Table.S5) across any two
   significant gene sets (Fig. [92]6A). Utilizing LASSO regression
   (Fig. [93]6B), candidate mRNA biomarkers were filtered from the
   intersection, yielding six mRNAs with optimal lambda values
   (Fig. [94]6C). We validated the expression levels of these mRNAs in
   peripheral blood samples using quantitative PCR (qPCR) (Fig. [95]6D).
   In the AMI group, ANKRD52, ART1, NRP2, and PPP1R15A expressions were
   elevated compared to the healthy control group, whereas BAIAP2L1 and
   CCNE1 were reduced. Notably, after treatment, these six mRNAs’
   expression levels did not significantly differ from those in the AMI
   group (Fig. [96]6E).

Fig. 6.

   Fig. 6
   [97]Open in a new tab

   Through various data analysis methods, significant gene sets were
   obtained in AMI, post-treatment of AMI, and healthy control group. (A)
   The UpSet plot illustrated the intersection and union relationships
   among DEGs (top 200), sPLS-DA component 1, component 2, tan module of
   WGCNA, and random forest (RF) top 200 genes. The ribbons between the
   dots represent the intersections between sets. Genes with an
   intersection count of two or more groups were selected for further
   analysis. (B) The extracted features were reduced via the LASSO
   regression. (C) LASSO coefficients of the variables. (D) The expression
   level of six mRNA in the AMI, Post-treatment of AMI and healthy
   controls. (E) Validation of 6 mRNA expression by fluorescence
   quantitative PCR.

Hub gene expression in the AMI mice model

   To further validate the relationship between AMI and the candidate
   biomarkers, we employed a mouse AMI model. Following immediate ligation
   of the left anterior descending (LAD) coronary artery, ECG recordings
   displayed a characteristic ST-segment elevation (Fig.S7A), indicating
   successful induction of myocardial ischemia. Histological analysis of
   myocardial sections revealed significant changes in the infarct area
   when compared to the sham group (Fig.S7B, C). Additionally, elevated
   serum cTnI levels quantitatively confirmed the occurrence of myocardial
   damage (Fig.S7 D). M-mode echocardiography was performed to evaluate
   cardiac structure and function (Fig.S7E). Detailed measurements of
   interventricular septum (IVS) thickness, left ventricular posterior
   wall (LVPW) thickness, left ventricular internal diameter (LVID),
   ejection fraction (EF), and fractional shortening (FS) revealed
   alterations in cardiac structure and contractility (Fig.S7F-J).
   Notably, the mRNA levels of ANKRD52 (Fig.S7K) and NRP2 (Fig.S7L) in
   peripheral blood were significantly elevated in the myocardial
   infarction model compared to the sham group, consistent with our
   findings in human data.

Discussion

   AMI is a significant contributor to cardiovascular mortality, and early
   diagnosis is crucial for effective treatment and improved outcomes.
   Although various biomarkers have been explored for the diagnosis of
   AMI, there is a continued need for reliable non-invasive biomarkers to
   accurately differentiate AMI patients from healthy individuals. This
   study aims to identify potential mRNA biomarkers in peripheral blood to
   aid in the early diagnosis and monitoring of AMI. In this study, our
   integrative analysis identified several candidate intersecting genes.
   ANKRD52, ART1, NRP2, and PPP1R15A were found to be upregulated, while
   BAIAP2L1 and CCNE1 demonstrated downregulation in the context of AMI.
   Members of the ANKR family are implicated in a diverse array of
   functions, including the formation of transcriptional complexes,
   initiation of immune responses, biogenesis and assembly of cation
   channels in membranes, and regulation of the cell cycle^[98]13–[99]15.
   ANKRD1 is identified as a cardiac-specific ankyrin repeat
   domain-containing protein principally expressed in the heart and
   implicated in the morphogenesis and function of cardiomyocytes^[100]16.
   Alterations in both the expression and phosphorylation levels of ANKRD2
   protein are known to mediate the balance between muscle physiology and
   pathological inflammatory responses^[101]17. Furthermore, ANKRD26 is
   associated with thrombopoiesis and the pathogenesis of autoimmune
   diseases. References 42–44 likely provide further substantiation for