| Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that significantly affects cognitive function. Early and accurate diagnosis is crucial for timely intervention, yet traditional diagnostic methods can be costly and invasive. This study explores a machine learning (ML) approach leveraging the Bio-Hermes dataset, which integrates blood-derived proteomics, neuropsychological test results, and demographic variables to improve the classification of individuals into cognitively normal (CN), mild cognitive impairment (MCI), and AD cohorts. We implement four ML models: Elastic Net, Classification and Regression Trees (CART), Random Forest, and Extreme Gradient Boosting (XGBoost). The results indicate that XGBoost achieves the highest accuracy (0.78) and a strong area under the receiver operating characteristic curve (AUC-ROC) (0.90), highlighting the potential of multimodal ML models in dementia classification. Additionally, the Mini-Mental State Exam (MMSE) and the proteomic marker pTau181 emerged as key predictive variables. This study underscores the feasibility of using blood-based proteomics in conjunction with cognitive assessments for early AD detection and advocates for further validation on larger, diverse cohorts. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.