Unraveling sex differences in Parkinson's disease through explainable machine learning

Area of Science:

• Neurology
• Machine Learning
• Genetics

Background:

• Sex differences significantly influence Parkinson's disease (PD) development and presentation.
• Current PD diagnostics and treatments often overlook these crucial sex-based distinctions.
• Existing research frequently prioritizes prevalence over detailed feature importance analysis in sex-specific PD studies.

Purpose of the Study:

• To develop an explainable machine learning (ML) model to identify and understand sex-specific factors in Parkinson's disease.
• To analyze feature importance and interactions, moving beyond simple prevalence rates for a deeper understanding of PD.
• To facilitate personalized medicine by tailoring data collection and analysis for male and female PD patients.

Main Methods:

• An explainable ML model was developed to integrate heterogeneous data (clinical, imaging, genetics, demographics).
• The model was used to identify sex-specific differences in PD diagnosis, predicting outcomes as 'Healthy' or 'Pathological'.
• Feature importance and interaction analyses were conducted to elucidate underlying factors influencing PD manifestation in males and females.

Main Results:

• The ML model identified muscle rigidity, autonomic and cognitive assessments, and family history as key PD diagnostic contributors, with notable sex differences.
• The genetic variant SNCA-rs356181 showed potentially greater significance in characterizing PD in males.
• Interaction analysis indicated a higher frequency of feature interplay among males compared to females.

Conclusions:

• Explainable ML provides critical insights into sex-specific PD pathophysiology, enhancing understanding beyond prevalence.
• Identified sex disparities can guide the development of tailored diagnostic biomarkers and therapeutic strategies for Parkinson's disease.
• The study underscores the need for sex-specific approaches in PD research, data collection, and clinical practice for improved patient outcomes.