Application of machine learning technology and miRNA biomarkers: A novel approach for epilepsy diagnosis

JOURNAL/atin/04.03/02274269-202512000-00001/figure1/v/2025-10-03T110833Z/r/image-tiff Epileptic patients experience recurrent seizures, which can lead to loss of consciousness or death. Recent studies have shown that microRNAs (miRNAs) play a crucial role in the onset of epilepsy. In this study, we developed machine learning models for diagnosing epilepsy using miRNA biomarkers. We used nucleotide sequence information and related target genes as biomarkers for each miRNA. This study utilized epilepsy-associated miRNAs and miRNAs from healthy individuals. We obtained a classification accuracy of 95.45% with the Naïve Bayes Multinomial classifier. We also used other classifiers, including Hoeffding Tree, Sequential Minimal Optimization, Random Tree, and Logistic Regression classifiers, which achieved reasonably high accuracies. These models were then tested on the epilepsy miRNAs from two independent datasets, and reasonable classification accuracies were observed. These findings indicate that the high performance of these models developed in this study suggests strong correlations between miRNA sequence information, target gene patterns, and their association with epilepsy. By combining our machine learning approach with a cost-effective miRNA testing strategy, this approach could be utilized as a novel method for epilepsy diagnosis.