Early Screening for Multiple Sclerosis Using Gut Microbiome and Machine Learning

Multiple Sclerosis is a chronic disease of the central nervous system that affects millions worldwide, and early detection is crucial for better treatment outcomes. Current detection methods are expensive and invasive, leading to a need for a cheaper and non-invasive screening tool. We utilized the gut microbiome and machine learning to develop a reliable screening tool for Multiple Sclerosis that can recommend further clinical examination. We identified optimal combinations of feature selection and machine learning algorithms as well as the associated biomarkers for Multiple Sclerosis in the gut microbiome. Our evaluation revealed that among various classifiers, Logistic Regression and Naive Bayes, particularly when trained on features filtered through Correlation-based Feature Selection and Information Gain methods, exhibited exceptional performance in terms of precision, recall, and F-measure consistently higher than 0.95 and high AUC values. Furthermore, our analysis of the biomarkers identified distinct microbial signatures between healthy and Multiple Sclerosis patients, notably the prevalence of certain taxa, such as genera Streptococcus and Dorea in Multiple Sclerosis patients, and phylum Bacteroides and species uniformis in healthy cases.