Subtypes of corneal neuropathy progression in Parkinson's disease and their association with clinical phenotypes

Parkinson’s disease (PD) is characterized by significant clinical heterogeneity, and there is an urgent need for biomarkers that can identify different progression subtypes. This paper utilizes a machine learning technique—Subtype and Stage Inference (SuStaIn)—which can reveal data-driven disease phenotypes with different temporal progression patterns from widely available cross-sectional patient studies. Corneal neuropathy is a promising peripheral biomarker for PD. Based on confocal microscopy of the cornea in PD patients, this study shows that the method can identify subgroups and their unique trajectories of regional neurodegeneration. SuStaIn can identify two subtypes merely through corneal data and uniquely characterize their temporal complexity. Further analysis revealed that these two subtypes correspond to different pathological protein transmission pathways respectively, and showed significant differences on key clinical symptom scales such as UPDRS-II, RBD-HK and FOGQ. The fine-grained patient stratification provided by SuStaIn has significantly enhanced the ability to predict the transformation between diagnostic categories, bringing new hope to the discovery of disease subtypes and precision medicine.