Related Experiment Videos
Parkinson's progression prediction using machine learning and serum cytokines.
Diba Ahmadi Rastegar1, Nicholas Ho1, Glenda M Halliday1
1ForeFront Dementia and Movement Disorders Laboratory, Brain and Mind Centre, Central Clinical School, Faculty of Medicine and Health, University of Sydney, Camperdown, NSW 2050 Australia.
NPJ Parkinson'S Disease
|August 3, 2019
Summary
Predicting Parkinson's disease (PD) progression is crucial. This study found that specific inflammatory markers in blood, measured using machine learning, can help forecast motor symptom severity in PD patients.
Area of Science:
- Neuroscience
- Immunology
- Biomedical Engineering
Background:
- Parkinson's disease (PD) exhibits heterogeneous symptom progression, complicating treatment and clinical trial analysis.
- Predictive models for PD progression are of significant interest to personalize patient care and improve research outcomes.
Purpose of the Study:
- To investigate the stability of inflammatory cytokines in PD patient serum over time.
- To develop machine learning models for predicting longitudinal clinical outcomes in PD using baseline serum cytokine measurements.
Main Methods:
- Serum samples from a longitudinally followed cohort of PD patients (with and without LRRK2 G2019S mutation) were analyzed.
- 27 inflammatory cytokines and chemokines were measured at baseline and after 1 year.
- Machine learning models were employed to predict clinical outcomes (Hoehn and Yahr scale, UPDRS III) using baseline cytokine data.
Main Results:
- The best prediction models achieved a normalized root mean square error (NRMSE) of 0.1123 for the Hoehn and Yahr scale and 0.1193 for the UPDRS III.
- Key predictors for motor symptom severity included macrophage inflammatory protein one alpha (MIP1α) and monocyte chemoattractant protein one (MCP1).
- Cytokine stability was assessed, providing insights into their longitudinal behavior in PD.
Conclusions:
- Peripheral inflammatory cytokines show potential utility in predicting Parkinson's disease progression.
- Machine learning models integrating cytokine data can aid in forecasting motor symptom severity in PD.
- Findings contribute to understanding peripheral inflammation dynamics in PD and its predictive value.