Author(s) :

Muchen Xu.

Volume/Issue :

Abstract :

Dementia, marked by cognitive decline and neuropsychiatric symptoms, significantly impacts individuals and society, especially with an aging global population. Despite the need for early diagnosis and intervention, current diagnostic methods are costly and invasive. This study investigates blood-based microRNAs (miRNAs) as non-invasive, cost-effective biomarkers for early dementia diagnosis and subtype differentiation. Using machine learning techniques, we analyzed serum miRNA expression profiles from the Gene Expression Omnibus database (GSE120584), which includes samples from dementia and cognitively normal controls. Our approach involved Support Vector Machine (SVM), Random Forest (RF), Recursive Feature Elimination (RFE), and Neural Networks for feature selection. Logistic Regression was used for classification. Pathway analysis was further performed on the target genes of the identified miRNA biomarkers to explore biological insights behind these biomarkers. We identified miRNAs such as miR-6777-3p, miR-1471, and miR-6806-5p as potential biomarkers for dementia diagnosis and miRNAs like miR-4290 and miR-3184-3p for subtype differentiation. Among the miRNA biomarkers, miR-371b-3p, miR-1539, and miR-4290 are newly discovered biomarkers that have not been mentioned in any studies before. Additionally, this study demonstrates the power of integrating deep learning with traditional machine learning techniques to find new outcomes. This study also reveals the connection between dementia and infectious diseases on a molecular level, providing new therapeutic insights in dementia.