Evaluation of miRNAs Involved in Cardiac Fibrosis: Approach Based on Machine Learning
نویسندگان
1 Department of Cardiology, School of Medicine, Zabol University of Medical Scienc-es, Zabol, Iran
2 School of Medicine, Jundishapur University of Medical Sciences, Ahvaz, Iran
3 Department of General Surgery, Lorestan University of Medical Science, Khorramabad, Iran
4 Department of Cardiology, Shariati Hospital, Tehran University of Medical Sciences, Tehran, Iran
5 Department of Cardiology, School of Medicine, Shahid Madani Hospital, Lorestan University of Medical Sciences, Lorestan, Iran
6 Department of Emergency Medicine, Faculty of Medicine, Ahvaz Jundishapur University of Medical Sciences, Ahvaz, Iran
7 Rajaie Cardiovascular Medical and Research Center, School of medicine, Iran University of Medical Sciences, Tehran, Iran
8 School of Medicine, Jundishapur University of Medical Sciences, Ahvaz, Iran
doi
10.22034/ircmj.2024.469622.1332چکیده
Background and Objectives: Defective cardiac function can lead to the excessive accumulation of extracellular matrix proteins in the heart. This study aimed to investigate and analyze the miRNAs involved in cardiac fibrosis using machine learning algorithms. Methods: A dataset related to the expression of miRNAs in healthy individuals and patients with cardiac fibrosis was collected from public sources and relevant clinical databases. We selected a study population of 50 individuals, comprising 25 healthy controls and 25 patients with cardiac fibrosis. Different machine learning algorithms, including support vector machines, random forests, and artificial neural networks were employed to analyze, categorize, and predict the role of miRNA expression changes in cardiac fibrosis. The validation of the machine learning model through both leave-one-out cross-validation and an independent dataset effectively supports the robustness of the results. The reported metrics for accuracy, sensitivity, and specificity were impressive. Results: A total of 78 miRNAs were found to be differentially expressed (adjusted P < 0.05) between the two groups, with 47 miRNAs upregulated and 31 miRNAs downregulated in the cardiac fibrosis group compared to the control group. The top 10 miRNAs selected by each method were compared, and a consensus set of 5 miRNAs (miR-21-5p, miR-29a-3p, miR-29c-3p, miR-30b-5p, and miR-133a-3p) was identified as the most informative features for distinguishing between cardiac fibrosis and control samples. Based on the results, the expression of miR-21-5p, miR-29a-3p, and miR-29c-3p was increased in cardiac fibrosis patients compared to the control group. On the contrary, it was found that the expression of miR-30b-5p and miR-133a-3p was increased in normal subjects compared to patients. The SVM algorithm with a radial basis function kernel was found to be the best-performing model, with an accuracy of 92%, sensitivity of 88%, specificity of 96%, and area under the receiver operating characteristic curve (AUC-ROC) of 0.95. Conclusion: In general, the evaluation of miRNAs in patients can be used as a biomarker for monitoring patients and applying treatment strategies.
