Integrating Machine Learning and Gene Expression Analysis for Dementia Subtype Classification and Biomarker Discovery

Dementia, a progressive neurodegenerative disorder affecting millions worldwide, poses significant challenges in early diagnosis and targeted treatment due to overlapping clinical symptoms. While gene expression analysis has provided molecular insights into dementia subtypes, the integration of machine learning (ML) remains underexplored. This study utilizes gene expression data from the GEO database and ML techniques to enhance dementia subtype classification and identify potential biomarkers. Rigorous preprocessing, including quantile normalization and SMOTE, ensured data integrity and minimized bias. Among all the models evaluated, the Support Vector Classifier (SVC) achieved highest accuracy (99%), while the Random Forest (RF) classifier proved valuable for feature selection. Key biomarkers, including MOBP, PDIA5, and RAB43, were identified, with pathway enrichment analysis highlighting subtype-specific molecular mechanisms such as axon guidance and disrupted protein processing. These findings demonstrate the potential of ML-driven approaches in dementia research, offering improved diagnostic accuracy and biomarker discovery. Future studies should focus on validating these biomarkers in independent cohorts and integrating multi-omics data to advance precision medicine for neurodegenerative disorders.