1st International and 10th National Iranian Conference on Bioinformatics
Home Accepted Papers
Machine learning-driven set of peripheral blood microRNAs as diagnostic biomarkers for myocardial infarction
Paper ID : 1341-ICB10
Authors:
Mehrdad Samadishadlou1, Zeynab Piryaei2, Farhad Bani3, Kaveh Kavousi *2
1Department of Medical Nanotechnology, Faculty of Advanced Medical Sciences, Tabriz University of Medical Sciences, Tabriz, Iran.
2University of Tehran
3a- Department of Medical Nanotechnology, Faculty of Advanced Medical Sciences, Tabriz University of Medical Sciences, Tabriz, Iran. b- Drug Applied Research Center, Tabriz University of Medical Sciences, Tabriz, Iran.
Abstract:
Cardiovascular disease is the leading cause of mortality worldwide and myocardial Infarction (MI) is responsible for 85% of cardiovascular disease mortality [1]. Since the survival rate in MI cases strongly depends on fast diagnosis and treatment, discovering novel biomarkers for rapid and accurate diagnosis is of great importance [2]. MicroRNAs are significant regulators of adaptive and maladaptive responses in cardiovascular diseases. Hence, microRNAs have undergone intensive research as possible therapeutically and diagnostic targets [3]. However, their role as novel biomarkers for diagnosing MI needs to be more investigated. The microarray GSE61741 [4] dataset has been downloaded from the Gene Expression Omnibus (GEO) database including 863 microRNAs expression profile in peripheral blood. The selected samples included 94 healthy (as the control) and 62 samples with MI (as the case). At the first, differentially expressed microRNAs has been identified using the limma package [5] with the adjusted P-value <0.05 and -1 > log2 FC > 1 criteria. Then, sequential forward and backward selection algorithms has been applied for feature selection. Finally, the support vector machine (SVM) algorithm [6] has been performed on selected microRNAs to classify samples with 10-fold cross-validation. 100 differentially expressed microRNAs has been identified in samples with MI compared to healthy samples. 35 microRNAs with the greatest importance has been selected using feature selection algorithms. Among them, a unique signature of five microRNAs (including hsa-miR-1246, hsa-miR-1258, hsa-miR-1279, hsa-miR-132*, and hsa-miR-142-3p) have been chosen and an SVM model has been trained with their expression values. The trained model predictive values are 0.92, 0.84, and 0.91, for AUC, sensitivity, and specificity, respectively. Based on our findings, the multi-marker approach increases predictive values in comparison to single microRNAs. Therefore, microRNA signatures derived from peripheral blood could be valuable novel biomarkers for more accurate diagnosis of MI.
Keywords:
Myocardial infarction, MicroRNA, Machine Learning, MicroRNA Signature, Biomarker
Status : Paper Accepted (Poster Presentation)