Advancements in Parkinson's Disease Prediction Using Machine Learning: A Neurological Perspective

Area of Science:

• Neuroscience
• Biochemistry
• Computational Biology

Background:

• Parkinson's disease (PD) is a progressive neurodegenerative disorder.
• Accurate prediction of PD severity is crucial for effective patient management and treatment planning.
• Current methods for assessing PD progression often rely on clinical evaluations, which can be subjective.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting Parkinson's disease severity.
• To integrate diverse data sources including cerebrospinal fluid proteomics, clinical assessments, and gait parameters.
• To assess the predictive performance of different machine learning algorithms, including ensemble methods.

Main Methods:

• A dataset of 248 Parkinson's disease patients was longitudinally monitored.
• Data included cerebrospinal fluid protein and peptide levels (227 proteins, 971 peptides), gait parameters, and Movement Disorder Society-sponsored revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) scores.
• Machine learning models employed included random forest, TensorFlow decision forests, phase-shift ensembling, linear regression, random forest regressor, decision tree regressor, and K-nearest neighbors.

Main Results:

• The custom phase-shift ensembling model achieved a superior average symmetric mean absolute percentage error (sMAPE) of 55 across all UPDRS sections.
• The random forest regressor demonstrated strong performance in predicting motor function severity (UPDRS-III), with an sMAPE of 77.32.
• These results indicate the models' effectiveness in capturing complex disease progression dynamics.

Conclusions:

• Integrating biological markers, clinical scores, and gait dynamics enables accurate modeling of PD progression.
• Ensemble-based approaches, particularly phase-shift ensembling, enhance prediction robustness and interpretability.
• This study underscores the value of multi-source data fusion and advanced machine learning for early PD prediction and treatment planning.