Hybrid Stacked Ensemble Regression Model for Predicting Parkinson’s Progression on Protein Data

Abstract

Parkinson’s Disease (PD) is a progressive neurological disorder marked by both motor and nonmotor symptoms. Accurate prediction of disease progression is critical for effective patient management. The research presents a Hybrid Stacked Ensemble Regression (HSER) model for predicting PD progression using protein and peptide data measurements, leveraging the Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (MDSUPDRS) scores. The researchers integrate three datasets: clinical data, protein data, and peptide data into a comprehensive feature-engineered dataset. The dataset is split into training and testing sets in four configurations for predicting the four UPDRS scores, namely updrs 1, updrs 2, updrs 3, updrs 4. The hybrid approach combines stacking and blending techniques. The researchers select ridge regression, gradient boosting, and extra trees as base models. A meta-model is trained using the algorithms’ out-of-fold estimates (ridge regression). The final predictions are obtained by averaging the predictions of the base models on the test data. The proposed HSER model exhibits enhanced performance compared to baseline models. These results underscore the promise of the hybrid model to enhance the prediction of PD progression, providing valuable insights for personalized treatment strategies. Future research can focus on refining model weights and exploring additional biomarkers to improve predictive accuracy.