Machine-Learning-Based Targeted Plasma Proteomic Analysis for Predicting Motor Progression in Parkinson's Disease: An...

Area of Science:

• Neuroscience
• Biomarkers
• Machine Learning

Background:

• Accurate prediction of Parkinson's disease (PD) motor progression is crucial for personalized medicine and clinical trials.
• Current methods for predicting PD progression are limited, posing a challenge for patient management.

Purpose of the Study:

• To develop and validate a machine-learning framework for stratifying Parkinson's disease patients based on motor progression risk.
• To integrate plasma protein data with clinical variables for enhanced predictive accuracy.

Main Methods:

• Analyzed baseline plasma samples from 211 early-stage PD patients using Olink proteomic assays.
• Employed machine learning algorithms, including Random Forest, to predict rapid versus slow motor progression (MDS-UPDRS Part III).
• Utilized SHapley Additive exPlanations (SHAP) for model interpretability and identified key protein predictors.

Main Results:

• The Random Forest model achieved an AUC of 0.751, outperforming clinical predictors alone (AUC 0.666).
• Integration of proteomic and clinical data improved prediction accuracy (AUC 0.773).
• Key predictors included Interleukin-6 (IL-6), Brain-Derived Neurotrophic Factor (BDNF), and Vascular Endothelial Growth Factor A (VEGF-A).

Conclusions:

• Targeted plasma protein profiling combined with interpretable machine learning offers a promising approach for PD motor progression risk stratification.
• This framework can aid in individualized patient counseling and optimize the selection of participants for disease-modifying clinical trials.