predALZ: An Ensemble Learning Framework for Identifying Genetic Biomarkers in Familial Alzheimer's Disease

Alzheimer's disease (AD) is a multifactorial neurodegenerative disorder with a substantial genetic contribution, especially in the early-onset form. Mutations in genes like APP, PSEN1, and PSEN2 serve as crucial biomarkers, indicating a heightened risk of developing AD. Leveraging these genetic markers, we introduce predALZ, a prediction model designed to enhance early detection of familial AD through genomic sequence analysis.

The model integrates data derived from genome-wide association studies (GWAS) and employs advanced feature encoding techniques to generate a robust representation of genomic patterns. A diverse ensemble of classifiers, namely XGBoost, Random Forest, LightGBM, and ExtraTrees, is employed to train the predALZ model, utilizing the generated feature vector for training.

The predALZ framework achieved 94% accuracy on an independent test set and approximately 96% accuracy in cross-validation for Alzheimer-related driver gene prediction. The ensemble model also yielded consistently high sensitivity, specificity, and Matthews correlation coefficient values, indicating stable and reliable classification performance.

The model's effectiveness was further rigorously validated through a comprehensive evaluation, using metrics such as accuracy, sensitivity, specificity, and Matthew's correlation coefficient. The study underscores the predictor's remarkable performance, consistently achieving 94% accuracy in an independent Test and ~96% in cross-validation.

These findings highlight predALZ's potential for application in predictive diagnostics and targeted therapeutic development for Alzheimer's disease.