Revised: April 3, 2026
Accepted: May 20, 2026
Published online: June 26, 2026
Processing time: 104 Days and 0.2 Hours
MicroRNAs (miRNAs) have emerged as key regulators and promising biomarkers in cardiovascular diseases (CVDs). To date, numerous miRNAs associated with CVDs have been identified. Machine learning (ML) models play a critical role in integrating multiple miRNAs into a unified predictive framework, thereby en
To synthesize the current evidence on the development and/or validation of ML-based diagnostic models using miRNA expression profiles for the classification and differentiation of CVDs, and to evaluate their diagnostic performance.
A literature search was conducted from inception to December 2025 across PubMed, the Cochrane Library, and ScienceDirect databases with specific key
The initial search identified 390 records, of which nine met the eligibility criteria and were included in this review. The included studies evaluated the diagnostic performance and clinical applicability of ML algorithms incorporating miRNA expression profiles across multiple CVD phenotypes, including acute coronary syndrome, acute myocardial infarction, coronary artery disease, essential hy
This review found that the integration of ML models and miRNA profiles for CVD diagnosis represents a promising strategy, with the potential to improve diagnostic accuracy through multivariable panels. However, the current evidence base remains limited, and further well-designed studies are needed to validate these models, standardize methodological approaches, and define their role in routine clinical diagnosis of CVDs.
Core Tip: Machine learning integration with microRNA profiles demonstrates promising discriminative performance (area under the curve-receiver operating characteristic often exceeding 0.80) for diagnosing and differentiating subtypes of cardiovascular diseases, including acute myocardial infarction, coronary artery disease, hypertension (HTN) subtypes, post-operative atrial fibrillation after coronary artery bypass grafting, and pulmonary HTN. However, the current evidence is limited by small-to-moderate sample sizes, predominant reliance on internal validation only, and lack of assessment of calibration, incremental value over established biomarkers (e.g., troponin), and standardization of pre-analytical and analytical methods. Large-scale prospective studies with external validation and rigorous standardization are essential before these approaches can be translated into routine clinical practice.
- Citation: Popat A, Sathipati S, Sharma P. Machine learning integration in microRNA-based markers for cardiovascular diseases: A systematic review. World J Cardiol 2026; 18(6): 120747
- URL: https://www.wjgnet.com/1949-8462/full/v18/i6/120747.htm
- DOI: https://dx.doi.org/10.4330/wjc.120747
MicroRNAs (miRNAs), first discovered in 1993, are small non-coding RNAs that function as critical post-transcriptional regulators of gene expression and have emerged as important mediators of cardiovascular pathobiology[1]. Their dysregulation has been implicated in the pathogenesis of a broad range of disease states, including cardiovascular diseases (CVDs)[2]. MiRNAs modulate the expression of specific target genes through translational repression or mRNA degradation, thereby influencing diverse cellular pathways and tissue- and cell-specific pathophysiological processes (Figure 1). Moreover, their biological role extends beyond the intracellular environment, as they are involved in intercellular signaling and diverse tissue- and cell-specific pathophysiological processes, including inflammation, fibrosis, apoptosis, vascular remodeling, and myocardial injury[2].
In CVDs, miRNAs have been associated with multiple clinical applications, including diagnosis, risk assessment, and prognostication. Among these, miR-223, miR-126, miR-21, and miR-150 have been most consistently investigated in relation to platelet function and platelet reactivity[3]. In other settings, circulating miRNAs, such as miR-21, miR-29a, miR-126, and miR-223, have shown promise as non-invasive biomarkers for cardiovascular risk stratification and prognostication in coronary artery disease (CAD), heart failure, and atrial fibrillation (AF)[4]. Ongoing investigations are evaluating the therapeutic potential of these miRNAs as molecular targets in a broad range of cardiovascular disorders, including atherosclerosis, AF, and other arrhythmias, heart failure, and acute coronary syndromes (ACSs)[5].
Machine learning (ML) methods are extensively employed in analyzing miRNAs and identifying specific miRNAs. The main advantage of ML in miRNA research is its ability to be adaptive and thus learn from the data presented[6,7]. Nevertheless, the development of robust ML models requires large, high-quality datasets. Before the advent of high-throughput sequencing technologies, the limited availability of comprehensive miRNA expression datasets represented a major barrier to the application of ML approaches. The advent of high-throughput sequencing technologies has generated large-scale datasets, thereby accelerating the integration of ML across biomedical and biological research[6,7].
While the use of miRNAs in CVDs has been increasing over the years, the use of ML-based methods in miRNA detection and validation is still in its infancy and has limited application[8]. This review was conducted to synthesize the current evidence on the development and/or validation of ML-based diagnostic models using miRNA expression profiles for the classification and differentiation of CVDs and to evaluate their diagnostic performance. Specifically, this review aimed to assess the diagnostic accuracy of ML models constructed using miRNAs for distinguishing among CVD subtypes and to evaluate their methodological rigor, including internal and external validation approaches.
This systematic review was conducted in accordance with the PRISMA 2020 statement[9]. The review protocol was prospectively registered in PROSPERO (CRD420251111151).
The literature search was conducted from inception to December 2025 across three databases, including the Cochrane Library, PubMed, and ScienceDirect, by two reviewers independently. These databases were selected as primary sources of peer-reviewed literature. Gray literature, such as preprints, conference proceedings/abstracts, dissertations, or unpublished reports, was not systematically searched, as the focus was on full-text, peer-reviewed primary studies with detailed methodological descriptions suitable for qualitative synthesis of ML model performance. The search was not extended to EMBASE, Web of Science, or Scopus due to institutional access limitations at the time of the review. The omission of these databases represents a limitation of the present study. The PubMed search utilized both MeSH terms and free-text keywords. The search strategy was adapted for each database to maximize sensitivity. The literature search method followed a systematic process carried out in sequential steps. In the first step, the authors defined the search criteria for the different databases used in the review. The step also entailed the definition of the keywords used in the different databases to search for the relevant articles. The keywords were used by combining two Boolean operators, “OR” and “AND”. The search string used was as follows: (“machine learning” OR “artificial intelligence”) AND (“microRNAs” OR “miRNA”) AND (“cardiovascular disease” OR “hypertension” OR “heart failure” OR “atherosclerosis” OR “acute coronary syndrome” OR “myocardial infarction”) and is detailed in online Supplementary material. These keywords were then modified according to the specifications of the different databases to maximize the search results. After the initial data search, the reviewers manually reviewed studies from the reference lists of relevant reviews and included studies.
After all the studies had been retrieved from the databases, two reviewers screened the studies according to the eligibility criteria. These criteria were predefined to minimize bias during the study selection process. The studies that fulfilled the inclusion criteria were included and sorted for data extraction. The inclusion criteria were studies reporting the development and/or validation of ML-based diagnostic models using miRNA expression profiles for the classification and differentiation of CVDs, studies that reported at least one quantitative diagnostic performance metric, studies that included patients with different CVDs, primary studies consisting of original research with primary outcomes, and studies that reported the different outcomes of applying ML-based approaches in investigating the various roles of miRNAs in CVDs with full-text availability in English. To maintain focus on the primary aim of diagnostic model performance, one biomarker-discovery-oriented study was also included because it explicitly evaluated the diagnostic potential of the identified miRNA signature and explored its application in a real-time diagnostic platform. This study was analyzed and discussed as a separate subgroup.
Additionally, the population, intervention, comparison, outcomes, and study design (PICOS) framework for the current review consisted of population = patients with CVDs, intervention = development and/or validation of ML-based diagnostic models using miRNA expression profiles for the classification and differentiation of CVDs, comparison = traditional (non-ML) methods for miRNA analysis or no comparator, outcomes = studies that reported quantitative diagnostic performance metric of developed and/or validated ML miRNA models, including accuracy, area under the curve (AUC)-receiver operating characteristic (ROC), sensitivity, specificity, and validation methodology (internal cross-validation and/or external validation) and study design = primary diagnostic accuracy studies including experimental and observational studies. However, the exclusion criteria involved studies that included patients with conditions other than CVDs, secondary studies, including meta-analyses, systematic reviews, letters to the editors, commentaries, and book chapters, studies that did not apply any ML or artificial intelligence (AI) models, and studies not reporting the required outcomes based on the current research objective.
Two independent reviewers conducted the data extraction process. Based on the agreement of both reviewers, the final data were included in the study. During the process, if the reviewers failed to reach a consensus, the discrepancy was handled by a third reviewer. Titles and abstracts were first screened to remove duplicates. Following this, the selected articles were checked to discard studies not fulfilling the PICOS criteria. Lastly, the remaining studies were assessed for full-text availability to confirm eligibility. Data were only extracted from those that met the review’s inclusion criteria. The extracted data from the studies consisted of the author details (last name and year), the setting, the sample size, the miRNAs identified, the role, and the study’s primary outcomes. The extracted data were formatted in a tabular format in an Excel spreadsheet (Microsoft Corporation, Redmond, Washington, United States).
For quality assessment, the current review adapted the QUADAS-2 tool, which includes four domains: Patient selection, index test, reference standard, and flow/timing. However, this tool is not ideal for quality assessment of studies investigating the use of ML models in diagnostic accuracy. Specifically, its primary limitation is that it does not account for AI-specific issues such as external validation, model overfitting, data curation, or algorithm training[10]. The enhanced QUADAS-AI tool, a version of QUADAS-2, is ideal for quality assessment of included studies for this study, but it is yet to be finalized[11]. Because QUADAS-2 does not fully capture AI/ML-specific issues, we added ML-specific signaling questions to the index test and flow/timing domains. These questions were based only on information reported in each included study and assessed internal validation, external validation, risk of overfitting, timing of miRNA/feature selection, class imbalance handling, calibration reporting, and transparency of model development. Internal validation was considered present if cross-validation, train-test splitting, or hold-out testing was reported. External validation was considered present if the model was tested in an independent cohort or dataset. These items were assessed qualitatively and incorporated into the overall QUADAS-2 judgment. This adapted approach is pragmatic and not a formally validated risk-of-bias tool.
The critical narrative technique was employed to integrate text, tables, and figures to summarize and validate evidence. Due to variations in the measured AI and ML outcomes, the current review could not conduct a meta-analysis. Therefore, all data in the included studies were qualitatively synthesized according to the Cochrane guidelines[12].
The initial search from databases reported 390 studies. Among these articles, 187 duplicates were identified and excluded during the assessment. The exclusion led to the screening of 203 articles based on title and abstract. Moreover, 135 irrelevant studies that did not fulfil the objective of the review were eliminated. Consequently, 68 articles with full-text versions were retrieved and rigorously evaluated for eligibility. Only nine of the 68 studies met the inclusion criteria, and 59 were excluded as follows: 16 did not include miRNAs, 13 did not include ML models, 12 did not include patients with CVDs, 4 were secondary studies, and 14 studies lacked information regarding the required outcomes. Hence a total of nine studies were included. The search strategy according to PRISMA flow diagram is presented in Figure 2.
This systematic review included nine studies conducted in different settings, including the United Kingdom, Germany, Iran, Luxembourg, and the United States[8,13-20]. The conditions investigated in the different studies included ACS, acute myocardial infarction (AMI), post-operative AF after coronary artery bypass grafting (CABG), hypertension (HTN) and pulmonary arterial HTN (PAH). The developed ML models and miRNAs were mainly applied to diagnose CVDs. While most studies developed and internally validated ML models for direct diagnostic classification, one study by Ren et al[13] focused primarily on ML-assisted biomarker discovery. ML was integrated with miRNA profiles in two primary ways across the studies: (1) As a tool for feature selection and biomarker discovery, primarily in Ren et al[13]; and (2) For developing and validating, internally and, in some cases, externally, diagnostic classification models in the remaining eight studies. The findings from the included studies are described in Table 1. All studies developed at least one ML model.
| Ref. | Setting | CVD | Sample size | Identified miRNAs | Role | Main outcome |
| Kayvanpour et al[8], 2021 | Germany | ACS | 66 ACS patients and 68 healthy controls; 148 suspected ACS patients initially enrolled | Top 10 miRNAs selected via ANOVA F-value: MiR-142-5p, miR-151a-3p, miR-145-5p, miR-186-5p, miR-191-5p, miR-29c-5p, miR-30d-5p, miR-342-5p, miR-362-5p, and miR-589-5p | Diagnosis of ACS | Machine learning models, including neural networks, classified ACS with high diagnostic performance |
| Ren et al[13], 2024 | United States | AMI/STEMI | 24 screening samples; n = 6 each for no known CAD, known CAD, STEMI-pre, and STEMI-PCI; validation samples also used | Already identified: MiR-499, miR-1, miR-208b. Newly identified: MiR-331-3p, miR-142-5p, miR-200b-3p, miR-132-3p, miR-3605-5p, miR-18a-5p, miR-423-5p, miR-543, miR-301a-3p | Diagnosis and differentiation of AMI/STEMI from stable CAD | SCAD/LASSO regularized LR identified a 9-miRNA profile that differentiated no known CAD, known CAD, STEMI-pre, and STEMI-PCI, with ROC curves approaching 1 in selected comparisons; explored for rapid point-of-care diagnosis using MIX.miR ion-exchange membrane technology |
| Samadishadlou et al[14], 2023 | Iran | AMI and stable CAD | Healthy (51), CAD (46), AMI (111) | Differentially expressed: Hsa-miR-21-3p, hsa-miR-32-3p, hsa-miR-186-5p. Additionally selected via AUC-ROC: Hsa-miR-197-5p, hsa-miR-29a-5p, hsa-miR-296-5p | Diagnosis of AMI; differentiating AMI from healthy samples and from CAD | Peripheral blood mononuclear cell-derived miRNA signatures were used to differentiate healthy controls, stable CAD, and MI samples |
| Samadishadlou et al[15], 2024 | Iran | AMI | Training set: 62 MI and 94 healthy controls; independent test set: 8 MI and 6 healthy controls | Hsa-miR-375-3p, hsa-miR-601, hsa-miR-34a-5p, hsa-miR-29c-5p, hsa-miR-330-5p, hsa-miR-199b-5p, hsa-miR-142-3p, hsa-miR-200a-3p, hsa-miR-132-5p, hsa-miR-133a-3p | Diagnosis of early-stage AMI | ML model identified 10 miRNAs with accuracy of 0.86 and AUC of 0.83 for diagnosing AMI |
| Reel et al[16], 2025 | United Kingdom | Essential HTN subtypes | Cushing’s syndrome (35), primary aldosteronism (109), paraganglioma/pheochromocytoma (75), primary HTN (111) | Hsa-miR-15a-5p, hsa-miR-32-5p, hsa-miR-485-3p, hsa-miR-495-3p, hsa-miR-1260a, hsa-miR-186-5p, hsa-miR-195-5p, hsa-miR-326, hsa-miR-139-5p, hsa-miR-133a-3p, hsa-miR-223-3p | Differentiation of endocrine HTN subtypes from primary HTN | Models trained with the miRNAs achieved balanced accuracy of 0.71-0.89 and AUCs of 0.8-0.9 in differentiating HTN subtypes and other conditions |
| Sajid et al[17], 2024 | Pakistan | CAD | CAD cases (58), controls without CAD/stenosis < 50% (55) | MiR-21, miR-33a, miR-133a, miR-145, miR-146a | Diagnosis of CAD | ML models using miRNA biomarkers showed good diagnostic performance for angiography-defined CAD |
| Yerukala Sathipati et al[18], 2025 | United States | Post-operative AF after CABG | Cases (7), controls (8) | Hsa-miR-19a-3p, hsa-miR-19b-3p, hsa-miR-184, hsa-let-7a-5p, hsa-miR-124-3p, hsa-miR-200a-3p, hsa-miR-423-5p, hsa-miR-96-5p, hsa-miR-100-5p, hsa-miR-17-5p | Prediction of post-operative AF after CABG | 10 pre-operative circulating miRNA signatures were used to develop ML models for predicting POAF after CABG |
| Jusic et al[19], 2023 | Luxembourg/Bosnia and Herzegovina | HTN | 89 cases, 85 controls | MiR-361-3p and miR-501-5p | Diagnosis of HTN | SVM model using the two miRNAs plus clinical characteristics achieved accuracy of 0.87, specificity of 0.91, sensitivity of 0.83, and AUC of 0.90 |
| Errington et al[20], 2021 | United Kingdom | PAH | 64 cases, 43 disease and healthy controls | MiR-636 and miR-187-5p | Diagnosis of PAH | Models using the two miRNAs showed high diagnostic accuracy in differentiating PAH patients from healthy controls |
The ML algorithms employed across the selected studies encompassed a diverse array, including regularized logistic regression (LR), support vector machines (SVM) incorporating linear, polynomial, and radial basis function (RBF) kernels, LR, random forest (RF) classifiers, gradient boosting machines (GB), k-nearest neighbors (kNN), XGBoost (XGB), decision trees, LogitBoost, simple logistic models, logistic model tree (LMT), sequential minimal optimization, CatBoost, AdaBoost (ABoost), artificial neural networks, regression partition tree (Rpart), and ensemble voting methods. Besides Ren et al[13], all the studies used the developed models for direct diagnostic classification of different diseases. However, Ren et al[13] used the ML models primarily to identify miRNAs specific to AMI and CAD. They then tested the identified miRNAs using a sensor platform to diagnose and differentiate between CAD and AMI. The characteristics of the different models developed are described in Table 2. The summary of key diagnostic performance metrics of ML models integrating miRNAs for CVD diagnosis is reported in Table 3.
| Ref. | Models evaluated | Internal validation strategy | External validation | Training dataset | Test/validation dataset | Performance metrics |
| Kayvanpour et al[8], 2021 | LR, kNN, LDA, NB, RF, CT, SVM, XGB, and ANN | The subjects were divided into training and test sets in the ratio of 9:1, respectively. This was repeated 10 times to enable ten-fold cross-validation | None | 90% of subjects; 121 samples per split | 10% of subjects; 13 samples per split | Accuracy, sensitivity, specificity, and ROC-AUC |
| Ren et al[13], 2024 | Regularized LR using either SCAD or LASSO | Leave-one-out cross-validation | None for the ML model; selected miRNAs were biologically evaluated in matched clinical samples using an ion-exchange membrane sensor platform | 24 subjects; 800-miRNA screening library (100%) | None; leave-one-out cross-validation was used because of small sample size | ROC curves and AUC (used to evaluate the selected miRNA combinations) |
| Samadishadlou et al[14], 2023 | A 2-layer architecture utilizing SVM (with linear, polynomial, and RBF kernels), LR, RF, kNN, GB, XGB, and DT models (layer 1 isolated healthy vs not-healthy; layer 2 separated MI vs CAD) | The data was split in a 7:3 ratio into the training and test sets, respectively. A ten-fold cross-validation followed this | None | 70% of all the samples | 30% of all the samples | AUC-ROC, accuracy, sensitivity, specificity, and confusion matrix |
| Samadishadlou et al[15], 2024 | SVM, GB, XGB and hard voting ensemble model | Done in 2 phases. In miRNA selection: The LASSO method was cross-validated using the dataset 10-fold to select the best miRNA to be used in model development. In model selection, the training dataset was split in a 7:3 ratio into training and validation datasets. The models were then cross-validated 5-fold on the datasets. The best-performing models were then tested on the independent dataset | Performed using an independent dataset (GSE29532) | GSE61741 (62 MI samples and 94 healthy samples) | GSE29532 (8 MI samples and 6 healthy samples) | Accuracy, AUC-ROC, sensitivity, and specificity |
| Reel et al[16], 2025 | J48, NB, IBk, RF, LB, LMT, SL, and SMO | The data was randomly split into training and testing sets in an 8:2 ratio for model development and validation | None | 80% of all the samples | 20% of all the samples | Balanced accuracy is the primary metric. Other metrics include sensitivity, specificity, AUC-ROC, F1 score and Kappa score |
| Sajid et al[17], 2024 | LR, SVM, nonlinear kNN, tree-based (DT, RF), GB, XGBM, CBoost, ABoost, and ensemble voting | The data subset was first split in an 8:2 ratio into a CV subset and a hold-out subset for final evaluation. The CV subset was then divided into 10 folds. Nine folds were used for training and one-fold for testing, and the process was repeated 10 times. The best models were then tested on the hold-out dataset | None | 80% of the 113 subjects (cohort: 58 CAD cases, 55 healthy controls) | 20% of the 113 subjects (hold-out subset) | Accuracy, sensitivity, specificity, AUC-ROC, performance evaluation measure, F-statistic, and P values |
| Yerukala Sathipati et al[18], 2025 | kNN, XGB, SVM, and RF | The data was split in a 8:2 ratio into training and validation datasets, respectively | Performed using an independent GEO dataset (GSE222739) | 80% of the dataset (n = 12) | 20% of the dataset (n = 3) | AUC-ROC, accuracy, specificity, and sensitivity |
| Jusic et al[19], 2023 | RF, SVM, MLP, XGB, kNN, Logit | Hyperparameter tuning used two repeated 10-fold CV. The final model was also evaluated using leave-one-out cross-validation | None | 147 subjects (89 from the validation cohort + 58 from the sequenced discovery cohort) | 23 subjects (randomly extracted as 20% of the 112-subject validation cohort) | AUC-ROC, balanced accuracy, F1 score, precision, sensitivity, and specificity |
| Errington et al[20], 2021 | RF, Rpart, LASSO, XGB, and Ensemble | The data was split into training and validation data sets. The models were then CV 10-fold in the training dataset | The models were externally validated using publicly available datasets | Two-thirds of the samples | One-third of the samples (validation set) | Sensitivity, specificity, AUC, correct classification rate (accuracy), positive predictive value, and negative predictive value |
| Ref. | Best model(s) | AUC-ROC (range or best) | Accuracy (best reported) | Sensitivity (best) | Specificity (best) | Notes on interpretation |
| Kayvanpour et al[8], 2021 | ANN was the best-performing model (SVM, kNN, LDA, and RF also performed highly) | 0.87-0.99 | 0.87-0.96 | 0.87-0.95 | 0.87-1.00 | Good internal discriminative performance, but no external validation; risk of optimistic bias |
| Ren et al[13], 2024 | Regularized LR (LASSO/SCAD) | 0.5 to approximately 1.0 | NR | NR | NR | Focus on miRNA identification; no full diagnostic model metrics |
| Samadishadlou et al[14], 2023 | Two-layer architecture utilizing SVM (RBF) | 0.96 (layer 2) to 1.0 (layer 1) | 0.96 (overall two-layer architecture) | 0.97 (layer 2) to 1.0 (layer 1) | 0.86 (layer 2) to 1.0 (layer 1) | Good internal performance (two-layer approach isolated healthy samples perfectly), but no external validation cohort utilized |
| Samadishadlou et al[15], 2024 | HVE (aggregating SVM, GB, and XGB) | 0.83 (HVE on test set) | 0.86 | 1.00 | 0.67 | Very small test set (14 samples total: 8 MI, 6 healthy) limits reliability; platform differences between training and test sets impacted individual model performance |
| Reel et al[16], 2025 | LMT/LogitBoost (along with SL and SMO) | 0.80-0.90 | 0.71-0.89 (balanced accuracy) | 0.43-0.95 | 0.83-1.00 | Moderate-large sample; balanced accuracy used |
| Sajid et al[17], 2024 | AdaBoost (for miRNA biomarkers) and GB (for atherosclerosis inflammatory biomarkers) | 0.88-0.95 (CV)/0.76-0.93 (hold-out) | 0.87-0.90 (CV)/0.78-0.96 (hold-out) | 0.88-0.92 (CV)/0.71-0.86 (hold-out) | 0.96-1.00 (CV)/0.81-1.00 (hold-out) | Moderate sample; strong internal metrics but no external validation |
| Yerukala Sathipati et al[18], 2025 | RF/XGB | 0.76-0.83 | 0.73-0.80 | 0.75-0.87 | 0.71 | Very small sample, high risk of overfitting, though external validation was performed |
| Jusic et al[19], 2023 | SVM | 0.90 | 0.87 | 0.83 | 0.91 | Moderate sample; internal only |
| Errington et al[20], 2021 | RF, XGB and Ensemble model | 0.82-0.85 | 0.81-0.83 | 0.86-0.91 | 0.64-0.71 | Study with external validation, more reliable estimates |
According to the QUADAS-2 tool, the primary concern in the included studies was reference tests for the diagnosis of CVDs (Table 4 and Figure 3). With the exception of Kayvanpour et al[8], none of the studies explicitly mentioned how the diagnosis of the patients was reached, or whether the tests used were the gold standard. ML-specific concerns include lack of external validation in 6/9 studies and small sample sizes in several investigations, which heighten risks of overfitting. A detailed study-by-study risk-of-bias assessment is provided in Table 4, incorporating both standard QUADAS-2 domains and explicit ML-related signalling questions.
| Ref. | Patient selection (risk of bias/applicability) | Index test (including feature selection and overfitting risk) | Reference standard | Flow and timing (including external validation) | Overall ML-specific concerns |
| Kayvanpour et al[8], 2021 | Low/Low | Low (10-fold CV was reported, but feature-selection timing was not fully clear) | Low (diagnosis strictly adjudicated by a board of 3 expert cardiologists using ESC guidelines) | Low (internal CV only; no external) | Moderate overfitting risk due to no external validation |
| Ren et al[13], 2024 | Unclear/Low | Low (LASSO/SCAD used for selection; validation on matched sample) | Unclear | High (no formal validation split reported) | Limited validation; ML mainly for feature identification |
| Samadishadlou et al[14], 2023 | Low/Low | Low (7:3 split and 10-fold CV) | Unclear | Low (internal only) | Small test set; class imbalance addressed using sample weighting |
| Samadishadlou et al[15], 2024 | Low/Low | Low (LASSO and 5/10-fold CV) | Unclear | Low (internal and independent test set) | Improved validation compared to prior work |
| Reel et al[16], 2025 | Low/Low | Low (8:2 split and balanced accuracy metric) | Unclear | Low (internal only) | Use of balanced accuracy helps with potential imbalance |
| Sajid et al[17], 2024 | Low/Low | Low (8:2 and 10-fold CV and hold-out) | Unclear | Low (internal only) | Ensemble methods; good internal practices |
| Yerukala Sathipati et al[18], 2025 | Unclear/Low (small n = 15) | Low (8:2 split) | Unclear | Low (external validation performed via independent GEO dataset) | Very small sample; high risk of overfitting |
| Jusic et al[19], 2023 | Low/Low | Low (8:2 and 10-fold CV) | Unclear | Low (internal only) | Small cohort |
| Errington et al[20], 2021 | Low/Low | Low (10-fold CV and external datasets) | Unclear | Low (external validation performed) | Strongest validation approach among included studies |
Due to heterogeneity in disease phenotypes, biospecimens, and analytic goals, the narrative synthesis is presented first by primary analytic goal (biomarker discovery vs diagnostic model development) and then by cardiovascular condition. All reported performance metrics should be interpreted cautiously, as the majority were obtained from internal validation in small-to-moderate cohorts, with external validation performed in only 3 studies. Specifically, Errington et al[20] validated their identified targets using an independent whole-blood RNA-sequencing dataset, while Yerukala Sathipati et al[18] externally validated their 10-miRNA panel utilizing an independent Gene Expression Omnibus (GEO) dataset (GSE222739).
Biomarker discovery-focused studies using ML: Ren et al[13] employed LR [least absolute shrinkage and selection operator (LASSO) or smoothly clipped absolute deviation] primarily to identify miRNAs differentially expressed between AMI and stable CAD from a large dataset. The study identified 9 candidate miRNAs (miR-331, miR-142, miR-200b, miR-132, miR-18a, miR-423, miR-3605, miR-543, and miR-301a) strongly associated with AMI. These were not used to train a full diagnostic classification model in the reported work; instead, the miRNA signature was explored for potential integration with a novel ion-exchange membrane technology aimed at real-time AMI diagnosis. This approach differs from the other included studies, which constructed and evaluated complete ML pipelines for classification.
Diagnostic ML model development and evaluation studies: CAD: One of the included studies, i.e., Sajid et al[17], included patients with CAD who presented to the hospital with angina. The study included 122 patients; however, 9 did not meet the inclusion criteria or refused to participate. Among the remaining patients, 58 were diagnosed with CAD, and 55 were controls without CAD but with less than 50% coronary stenosis. In the study, 10 ML models were developed and trained using the existing datasets. The best model with the respective miRNA combination was then validated. The best three models were ABoost modelled on three miRNAs, i.e., (mR-33a, mR-133a, and mR-146a; M1), ABoost modelled on five miRNAs, (mR-33a, mR-21, mR-133a, mR-146a, and mR-145; M2), and RF modelled on three miRNAs (mR-33a, mR-145, and mR-146a; M3). Two of the models achieved very high AUC-ROC in the detection of CAD, i.e., 0.93 for M1 and 0.97 for M2. The third model, M3, had an AUC-ROC of 0.79. All the models, however, had good diagnostic accuracy, with M1 and M2 having a diagnostic accuracy of 96% and M3 having an accuracy of 93%. Lastly, the sensitivity of M1, M2, and M3 was 86%, 100% and 71% respectively, while the specificity was 100%, 94% and 88% respectively. Given the moderate sample size and lack of independent external validation, these promising findings should be interpreted cautiously and require confirmation in larger, independent cohorts.
ACS: Four of the included studies investigated the use of ML methods in the diagnosis of ACS using miRNA[8,13-15]. One study included patients with ACS, including those with angina and myocardial infarction (MI), and the other three focused on MI (often comparing it to stable CAD or healthy controls). Kayvanpour et al[8] used 34 different miRNAs to construct 9 ML models to diagnose ACS (Table 2). All the models had an AUC ranging from 0.87 to 0.99, indicating that the models had good discriminative ability in differentiating ACS from non-ACS patients. Similarly, the models demonstrated reasonable specificity, ranging from 0.87 to 1.00. However, while the best ML model reached an accuracy of 0.96 and a sensitivity of 0.95, the lower bounds of the extracted metrics (accuracy of 0.72 and sensitivity of 0.42) actually reflect the performance of the standard single-point high-sensitivity troponin T test at a higher cut-off, rather than the ML models[8]. With a moderate sample and repeated 10-fold cross-validation (9:1 train-test split), models showed good discriminative ability (AUC: 0.87-0.99). However, the lack of external validation means performance may be overestimated.
Ren et al[13] used ML methods to identify miRNAs and then explored their application in differentiating between AMI and CAD. The study then identified 9 miRNAs, i.e., miR-331, miR-142, miR-200b, miR-132, miR-18a, miR-423, miR-3605, miR-543, and miR-301a, which were found to be strongly associated with AMI.
Samadishadlou et al[14,15] constructed ML models using miRNAs (Table 2). Using ML models, Samadishadlou et al[14] identified single miRNAs that had good diagnostic accuracy for MI. Further evaluation of the discriminative ability of these miRNAs to separate AMI from CAD found AUC-ROC as follows: Hsa-miR-186-5p had an AUC of 0.86, hsa-miR-197-5p had an AUC of 0.84, hsa-miR-29a-5p had an AUC of 0.83, hsa-miR-21-3p had an AUC of 0.85, hsa-miR-296-5p had an AUC of 0.80, and hsa-miR-32-3p had an AUC of 0.70. Furthermore, the authors tested models trained using these selected miRNAs. The best model was SVM, which had good AUC and accuracy in differentiating CAD and AMI. The SVM linear model had an accuracy of 0.82 and the best AUC of 0.93 in the test set, while the SVM-RBF model had an accuracy of 0.94 and the best AUC of 0.96 in the test set. The 7:3 train-test split plus 10-fold CV yielded SVM models with strong test-set AUC (up to 0.96) for differentiating AMI from CAD. The relatively larger and more balanced dataset strengthens confidence in these findings compared to smaller studies.
In contrast, Samadishadlou et al[15] utilized LASSO + ML models which achieved AUC 0.78-0.85 on their validation set. The use of an external test set is a strength, but the tiny test cohort (n = 14) raises concerns about reliability and overfitting. They also used a similar approach with different datasets for training and testing. Among the identified miRNAs, hsa-miR-200a-3p had the highest AUC of 0.76. Among the developed models on the internal validation set, SVM had the best AUC-ROC of 0.85, followed by XGB (AUC = 0.82) and GB (AUC = 0.78), though a hard voting ensemble ultimately achieved the best AUC of 0.83 on the independent test set. Overall, AMI studies with larger or publicly validated datasets (Samadishadlou et al[14,15], Kayvanpour et al[8]) reported more consistent high AUCs than very small cohorts. While Kayvanpour et al[8] and Samadishadlou et al[14] relied solely on internal validation, Samadishadlou et al[15] successfully utilized an independent external dataset. Although models showed high discriminative performance in internal validation, these findings were derived from moderate sample sizes and, for most of the other studies, lack external validation, meaning they should therefore be interpreted cautiously.
Jusic et al[19] investigated the utility of the ML model based on miRNAs in differentiating between patients with HTN and healthy controls. The SVM model was then developed using the most commonly identified miRNAs, i.e., miR-361-3p and miR-501-5p, alongside five clinical risk factors. The developed model was able to classify HTN with an accuracy of 0.87, a specificity of 0.91, a sensitivity of 0.83, and an AUC of 0.90[19].
Reel et al[16] expanded the scope further and investigated the utility of miRNA profiles in identifying HTN subtypes. Similar to the other studies, this study identified miRNAs to be used in classification and constructed ML models using these miRNAs. The most prominent miRNAs in the study were hsa-miR-32-5p and hsa-miR-15a-5p. The LMT model achieved an accuracy of 0.89 and an AUC of 0.9 in differentiating primary HTN from endocrine HTN, and an accuracy of 0.73 and AUC of 0.8 in differentiating primary HTN from primary aldosteronism. On the other hand, the logit boost model achieved an AUC of 0.9 in differentiating between primary HTN and Cushing’s syndrome, pheochromocytoma, or paraganglioma. Models achieved AUC = approximately 0.90 and accuracy > 0.80 for classifying HTN or its subtypes. These comparatively larger cohorts provide somewhat more robust estimates than tiny-sample studies, although external validation is still absent. Models achieved high AUC in internal validation; however, given the moderate sample sizes and lack of external validation, these results require confirmation in independent cohorts.
One of the studies, Yerukala Sathipati et al[18], first used PCR to identify circulating miRNAs that could be used in the diagnosis of AF. They identified 84 miRNAs and then selected the top 10 using a correlation-based feature selection method, which were then used in developing the models. They used a very small case-control dataset (7 post-operative atrial fibrillation cases, 8 controls; total n = 15) with pre-operative circulating miRNAs. Among the selected identified miRNAs in the analysis were hsa-miR-19b-3p, hsa-miR-19a-3p, hsa-miR-124-3p, hsa-let-7a-5p, hsa-miR-184, hsa-miR-423-5p, hsa-miR-200a-3p, hsa-miR-100-5p, hsa-miR-96-5p, and hsa-miR-17-5p. Four predictive algorithms namely, kNN, XGB, SVM, and RF were constructed as part of the investigation. Evaluation on the testing data revealed the following performance metrics: The RF model achieved an accuracy of 0.80, AUC of 0.76, specificity of 0.71, and sensitivity of 0.87; the kNN model recorded an AUC of 0.77, accuracy of 0.80, sensitivity of 0.62, and specificity of 1.00; the XGB model demonstrated a sensitivity of 0.75, specificity of 0.71, AUC of 0.83, and accuracy of 0.73; and the SVM model showed an AUC of 0.60, sensitivity of 0.37, specificity of 0.57, and accuracy of 0.46. Overall, the SVM algorithm exhibited the poorest diagnostic capability. Given the extremely small sample size, these results carry a high risk of overfitting and should be interpreted very cautiously, despite the authors utilizing an independent GEO dataset for external validation; they are hypothesis-generating rather than ready for clinical application.
Errington et al[20], developed four ML models (RF, Rpart, LASSO, and XGB) using 20 miRNAs, among which two (miR-187-5p and miR-636) were used across all the 4 models. The RF model was developed using 10 miRNAs and had a specificity of 0.71, a positive predictive value (PPV) of 0.83, sensitivity of 0.86, a negative predictive value (NPV) of 0.77 and an AUC of 0.84 in classifying PAH. The Rpart was developed using 4 miRNAs and achieved a specificity of 0.64, a PPV of 0.80, sensitivity of 0.91, an AUC of 0.79, and a NPV of 0.82 in classifying PAH. The LASSO model was developed using 13 miRNAs and achieved a specificity of 0.64, a PPV of 0.77, an AUC of 0.79, sensitivity of 0.77, and a NPV of 0.64 in classifying PAH. Lastly, the XGB model was developed using 8 miRNAs and achieved a sensitivity of 0.91, a specificity of 0.71, a PPV of 0.83, a NPV of 0.83 and an AUC of 0.82 in classifying PAH[20]. The inclusion of external validation makes these findings more credible and less prone to overfitting compared to most other included studies. Further validation in larger prospective cohorts is still warranted.
CVDs are the most prevalent diseases and the most common cause of mortality among non-communicable diseases[21]. Their significant health burden necessitates a timely diagnosis and treatment to optimize patient outcomes. Studies have shown that in CVD, such as AMI, the changes in miRNA expression have a significant role in the disease’s progression[22]. This has therefore made miRNAs a good target as non-invasive biomarkers of CVDs. Hence, to the best of our knowledge, this is one of the first systematic reviews to investigate the integration of ML with miRNA expression profiles for the diagnosis and differentiation of various CVD subtypes. The included studies were heterogeneous in terms of targeted CVD phenotypes (e.g., acute vs chronic conditions), biospecimen sources, and primary analytic goals. Most studies developed end-to-end ML classification models for diagnosis or differentiation, while Ren et al[13] focused on ML-driven miRNA discovery to support a novel real-time diagnostic platform. This heterogeneity, although reflective of the emerging field, limits direct comparability and underscores the need for standardized approaches in future research.
Previous studies have already established that miRNA expression is significantly dysregulated in CVDs as miRNA-155 levels are reduced considerably in CAD and AMI[23]. Extensive research has identified many miRNAs with different roles in CVDs. However, the main limitation of current research is the continuous discovery of new miRNAs, with limited exploration of the role of each of the miRNAs. The main advantage of ML integration is its ability to handle large datasets, thus increasing the possibility of evaluating different miRNAs and reevaluating them in different studies.
There are different ways in which ML is integrated into miRNA research in CVDs. The first application of ML models is to identify patterns of miRNAs in different CVDs. In this review, one of the studies, i.e., Ren et al[13], only applied ML to identify the miRNAs before validating their efficacy using an ion-exchange membrane. The ML models enabled them to identify nine miRNAs with good discriminative ability for AMI[13].
The other utility of ML models is in direct diagnosis and detection of different CVDs using different miRNA profiles. The second utility is more comprehensive. The extensiveness is because it first entails using the ML models to identify the types of miRNA that are differentially expressed in different CVDs. The identified miRNAs are then used to construct ML models for diagnosing different CVDs and mitigating the inherent limitation of using individual miRNAs, which have been shown to have lower discriminative ability for closely related CVDs such as CAD and AMI[14]. The developed models have been shown to have high AUCs in differentiating different CVDs and even in differentiating healthy and unhealthy samples[8].
Despite these promising discriminative metrics (AUC-ROC), several critical issues must be addressed before miRNA-ML approaches can be considered for routine clinical use. Most studies reported only discrimination performance with limited or no assessment of model calibration (i.e., agreement between predicted probabilities and observed outcomes), which is essential for reliable clinical decision-making. Reproducibility remains uncertain due to small sample sizes, heterogeneity in biospecimens (plasma, serum), and variability in miRNA quantification platforms.
Important pre-analytical and analytical factors were rarely addressed, including batch effects, standardization of RNA extraction and normalization methods, and pre-analytical variation (e.g., impact of hemolysis, sample storage conditions, or timing of blood draw after the acute event). These factors can substantially influence circulating miRNA levels and limit generalizability across settings. Biological plausibility of the selected miRNA panels was variably discussed; while certain miRNAs (such as members of the miR-133 and miR-208 families) have established links to cardiac injury and remodeling, many others require further functional validation studies.
Furthermore, with the exception of Kayvanpour et al[8] and Errington et al[20], none of the included studies evaluated the incremental value or added clinical utility of miRNA-ML models over established diagnostic tools, most notably high-sensitivity cardiac troponin for AMI (compared in Kayvanpour et al[8]) or natriuretic peptides for PAH (compared in Errington et al[20]). A clinically useful diagnostic biomarker panel must demonstrate meaningful improvement in diagnostic accuracy, risk reclassification, or decision-making beyond current standard-of-care tests. The absence of such analyses represents a critical gap and is one of the main reasons why translation of these promising miRNA-ML approaches into routine clinical practice remains premature.
Additional ML-specific barriers include the predominance of internal validation only (external validation performed in only three studies), risk of overfitting, potential feature selection leakage, and lack of prospective real-world testing. Assay standardization, regulatory approval pathways for multiplex miRNA tests, and seamless integration into existing diagnostic workflows also represent substantial hurdles.
While this study applied a robust approach in identifying all the current evidence on integrating ML models in miRNA use in CVDs, it had some limitations. One of the limitations is that it included a limited number of studies, which limited the generalizability of the results. Moreover, these studies were highly heterogeneous regarding the ML approach used and the miRNAs investigated. Such a limitation, therefore, hindered the pooling of the reported outcomes. We could therefore not use statistical evidence to draw firm conclusions on the topic. Secondly, apart from Errington et al[20], Yerukala Sathipati et al[18], and Samadishadlou et al[15], all the other included studies lacked external validation of the models developed. Due to a lack of external validation, the utility of these models in different clinical settings cannot be determined. This predominance combined with generally small-to-moderate sample sizes across the nine studies, increases the risk of overfitting and optimistic bias. Lack of external validation indicates that the integration of ML models in miRNA research is still in its infancy and necessitates further research to ensure its integration into clinical practice. Consequently, descriptors such as “high diagnostic accuracy” or “good discriminative ability” in the original studies must be interpreted cautiously; apparent strong performance in development or internal test sets may not hold in independent, real-world cohorts. With the exception of Kayvanpour et al[8] and Errington et al[20], no study assessed the incremental diagnostic value of the miRNA-based ML models beyond existing clinical biomarkers and risk scores.
Although the current review searched three major databases (PubMed, Cochrane Library, and ScienceDirect), it did not include EMBASE, Web of Science, or Scopus due to institutional access limitations. A broader set of search terms such as “deep learning”, “neural network”, “support vector machine”, “random forest”, and “classification model” may have improved retrieval.
This review found that integrating ML models and miRNAs in diagnosing CVDs is a promising approach with the potential of offering improved accuracy through multivariate panels. However, the evidence is still limited, and different milestones are yet to be reached. Hence, the review recommends future studies to validate their models externally, assess calibration and incremental value, to provide evidence on the utility of these ML models beyond the academic and research setting. Additionally, there is a need to conduct studies at a large-scale in larger sample sizes, which will increase the robustness of the developed models. Moreover, models should be developed on already identified miRNAs to enable their validation in different settings and thus help in assessing the effectiveness of these already identified miRNAs, and therefore increase the chances of the integration of these models in clinical practice.
| 1. | Lee RC, Feinbaum RL, Ambros V. The C. elegans heterochronic gene lin-4 encodes small RNAs with antisense complementarity to lin-14. Cell. 1993;75:843-854. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 10233] [Cited by in RCA: 8902] [Article Influence: 269.8] [Reference Citation Analysis (6)] |
| 2. | Wronska A. The Role of microRNA in the Development, Diagnosis, and Treatment of Cardiovascular Disease: Recent Developments. J Pharmacol Exp Ther. 2023;384:123-132. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 24] [Cited by in RCA: 37] [Article Influence: 12.3] [Reference Citation Analysis (0)] |
| 3. | Pedersen OB, Grove EL, Kristensen SD, Nissen PH, Hvas AM. MicroRNA as Biomarkers for Platelet Function and Maturity in Patients with Cardiovascular Disease. Thromb Haemost. 2022;122:181-195. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 6] [Cited by in RCA: 17] [Article Influence: 3.4] [Reference Citation Analysis (0)] |
| 4. | Searles CD. MicroRNAs and Cardiovascular Disease Risk. Curr Cardiol Rep. 2024;26:51-60. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 53] [Reference Citation Analysis (0)] |
| 5. | Iacobescu L, Ciobanu AO, Corlatescu AD, Simionescu M, Iacobescu GL, Dragomir E, Vinereanu D. The Role of Circulating MicroRNAs in Cardiovascular Diseases: A Novel Biomarker for Diagnosis and Potential Therapeutic Targets? Cureus. 2024;16:e64100. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 3] [Reference Citation Analysis (0)] |
| 6. | García-Fonseca Á, Martin-Jimenez C, Barreto GE, Pachón AFA, González J. The Emerging Role of Long Non-Coding RNAs and MicroRNAs in Neurodegenerative Diseases: A Perspective of Machine Learning. Biomolecules. 2021;11:1132. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 28] [Cited by in RCA: 35] [Article Influence: 7.0] [Reference Citation Analysis (0)] |
| 7. | Daniel Thomas S, Vijayakumar K, John L, Krishnan D, Rehman N, Revikumar A, Kandel Codi JA, Prasad TSK, S S V, Raju R. Machine Learning Strategies in MicroRNA Research: Bridging Genome to Phenome. OMICS. 2024;28:213-233. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 1] [Cited by in RCA: 3] [Article Influence: 1.5] [Reference Citation Analysis (0)] |
| 8. | Kayvanpour E, Gi WT, Sedaghat-Hamedani F, Lehmann DH, Frese KS, Haas J, Tappu R, Samani OS, Nietsch R, Kahraman M, Fehlmann T, Müller-Hennessen M, Weis T, Giannitsis E, Niederdränk T, Keller A, Katus HA, Meder B. microRNA neural networks improve diagnosis of acute coronary syndrome (ACS). J Mol Cell Cardiol. 2021;151:155-162. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 2] [Cited by in RCA: 18] [Article Influence: 3.0] [Reference Citation Analysis (0)] |
| 9. | Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, Shamseer L, Tetzlaff JM, Akl EA, Brennan SE, Chou R, Glanville J, Grimshaw JM, Hróbjartsson A, Lalu MM, Li T, Loder EW, Mayo-Wilson E, McDonald S, McGuinness LA, Stewart LA, Thomas J, Tricco AC, Welch VA, Whiting P, Moher D. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 9803] [Reference Citation Analysis (0)] |
| 10. | Guni A, Sounderajah V, Whiting P, Bossuyt P, Darzi A, Ashrafian H. Revised Tool for the Quality Assessment of Diagnostic Accuracy Studies Using AI (QUADAS-AI): Protocol for a Qualitative Study. JMIR Res Protoc. 2024;13:e58202. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 35] [Reference Citation Analysis (0)] |
| 11. | Sounderajah V, Ashrafian H, Rose S, Shah NH, Ghassemi M, Golub R, Kahn CE Jr, Esteva A, Karthikesalingam A, Mateen B, Webster D, Milea D, Ting D, Treanor D, Cushnan D, King D, McPherson D, Glocker B, Greaves F, Harling L, Ordish J, Cohen JF, Deeks J, Leeflang M, Diamond M, McInnes MDF, McCradden M, Abràmoff MD, Normahani P, Markar SR, Chang S, Liu X, Mallett S, Shetty S, Denniston A, Collins GS, Moher D, Whiting P, Bossuyt PM, Darzi A. A quality assessment tool for artificial intelligence-centered diagnostic test accuracy studies: QUADAS-AI. Nat Med. 2021;27:1663-1665. [RCA] [PubMed] [DOI] [Full Text] [Cited by in Crossref: 73] [Cited by in RCA: 162] [Article Influence: 32.4] [Reference Citation Analysis (0)] |
| 12. | McKenzie JE, Brennan SE. Synthesizing and presenting findings using other methods. In: Higgins JPT, Thomas J, Chandler J, Cumpston M, Li T, Page MJ and Welch VA, editors. Cochrane Handbook for Systematic Reviews of Interventions. Hoboken: Wiley, 2019. [DOI] [Full Text] |
| 13. | Ren X, Zhou R, Ronan G, Ozcebe SG, Ji J, Senapati S, March KL, Handberg E, Anderson D, Pepine CJ, Chang HC, Liu F, Zorlutuna P. Towards real-time myocardial infarction diagnosis: a convergence of machine learning and ion-exchange membrane technologies leveraging miRNA signatures. Lab Chip. 2024;24:5203-5214. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 7] [Reference Citation Analysis (0)] |
| 14. | Samadishadlou M, Rahbarghazi R, Piryaei Z, Esmaeili M, Avcı ÇB, Bani F, Kavousi K. Unlocking the potential of microRNAs: machine learning identifies key biomarkers for myocardial infarction diagnosis. Cardiovasc Diabetol. 2023;22:247. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 14] [Reference Citation Analysis (0)] |
| 15. | Samadishadlou M, Rahbarghazi R, Kavousi K, Bani F. An exploration into the diagnostic capabilities of microRNAs for myocardial infarction using machine learning. Biol Direct. 2024;19:127. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 5] [Cited by in RCA: 5] [Article Influence: 2.5] [Reference Citation Analysis (0)] |
| 16. | Reel S, Reel PS, Van Kralingen J, Larsen CK, Robertson S, MacKenzie SM, Riddell A, McClure JD, Lamprou S, Connell JMC, Amar L, Pecori A, Tetti M, Pamporaki C, Kabat M, Ceccato F, Kroiss M, Dennedy MC, Stell A, Deinum J, Mulatero P, Reincke M, Gimenez-Roqueplo AP, Assié G, Blanchard A, Beuschlein F, Rossi GP, Eisenhofer G, Zennaro MC, Jefferson E, Davies E. Identification of hypertension subtypes using microRNA profiles and machine learning. Eur J Endocrinol. 2025;192:418-428. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 6] [Reference Citation Analysis (0)] |
| 17. | Sajid M, Hassan A, Khan DA, Abbas Gilani ST, Khan SA and Bakhshi AD. AI-Based Early Detection of Coronary Artery Disease Using Atherosclerosis Inflammatory and MicroRNA Biomarkers in Angina Patients. 2024 International Conference on Robotics and Automation in Industry (ICRAI); Rawalpindi, Pakistan. New York: IEEE, 2024. [DOI] [Full Text] |
| 18. | Yerukala Sathipati S, Carter T, Soodi D, Somto N, Shukla SK, Petronovich J, Ingrid G, Braxton J, Sharma P. MicroRNA signature predicts post operative atrial fibrillation after coronary artery bypass grafting. Sci Rep. 2025;15:18658. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in RCA: 4] [Reference Citation Analysis (0)] |
| 19. | Jusic A, Junuzovic I, Hujdurovic A, Zhang L, Vausort M, Devaux Y. A Machine Learning Model Based on microRNAs for the Diagnosis of Essential Hypertension. Noncoding RNA. 2023;9:64. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 6] [Reference Citation Analysis (0)] |
| 20. | Errington N, Iremonger J, Pickworth JA, Kariotis S, Rhodes CJ, Rothman AM, Condliffe R, Elliot CA, Kiely DG, Howard LS, Wharton J, Thompson AAR, Morrell NW, Wilkins MR, Wang D, Lawrie A. A diagnostic miRNA signature for pulmonary arterial hypertension using a consensus machine learning approach. EBioMedicine. 2021;69:103444. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 29] [Cited by in RCA: 51] [Article Influence: 10.2] [Reference Citation Analysis (0)] |
| 21. | World Health Organization. Noncommunicable diseases. Sep 25, 2025. [cited 6 March 2026]. Available from: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases. |
| 22. | Laggerbauer B, Engelhardt S. MicroRNAs as therapeutic targets in cardiovascular disease. J Clin Invest. 2022;132:e159179. [RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)] [Cited by in Crossref: 1] [Cited by in RCA: 113] [Article Influence: 28.3] [Reference Citation Analysis (0)] |
| 23. | Kondracki B, Kłoda M, Jusiak-Kłoda A, Kondracka A, Waciński J, Waciński P. MicroRNA Expression in Patients with Coronary Artery Disease and Hypertension-A Systematic Review. Int J Mol Sci. 2024;25:6430. [RCA] [PubMed] [DOI] [Full Text] [Cited by in RCA: 9] [Reference Citation Analysis (0)] |