Multiple Sclerosis Biomarkers Detection by a BiLSTM Deep Learning Model for miRNA Data Analysis

Abstract

High-throughput data technology has enabled studies on microRNA analysis to evolve in the field of early disease diagnosis. Multiple Sclerosis is one of the most known chronic autoimmune diseases that can cause severe disabilities, including partial blindness, motor disabilities, and considerable psychological impact. This work introduces a complete BiLSTM deep learning model for analyzing miRNA data of Multiple Sclerosis patients to provide early detection for this disease. The introduced model is based on a preprocessing flow published earlier by the authors. The experiments were conducted on a dataset of 215 transcriptomic miRNA samples of treated and untreated Multiple Sclerosis patients. The implicated results were quite promising, as the produced sensitivity, specificity, precision, accuracy, and F1 scores of (0.785,0.789,0.788, 0.8, and 0.79) respectively, were achieved. To ensure model robustness, the obtained accuracy of the introduced model was compared to two other state-of-art models, and the proposed BiLSTM has relatively outperformed the other literature models.