Investigating Molecularly Defined Clusters of Parkinson’s Disease Based on Multi-Omics Data With Clinical and Biological Implications

Abstract

Introduction: Parkinson’s disease (PD) is the most prevalent neurodegenerative movement disorder, involving motor and cognitive symptoms that are heterogeneous across individuals. Current clinical tests for assessing PD often miss important information about molecular and biological heterogeneity in the disease. To address this problem, we sought to answer the question of whether molecular data could provide generalizable clusters of PD with clinical and biological implications. We hypothesized that the molecular data would provide more informative classifications of PD relative to clinical tests alone. Methods: We performed a clustering analysis with bulk transcriptomic data from whole blood and proteomic data from urine and cerebrospinal fluid from the Parkinson’s Progression Markers Initiative. In this analysis, we used relevant features that correlated with longitudinal PD severity. We analyzed the resulting clusters for clinical and biological implications, considering the Unified PD Rating Scale (UPDRS), the Montreal Cognitive Assessment, dopamine transporter scans, seed amplification assays, gene set enrichment, and cell-type deconvolution. To assess generalizability, we created logistic regression and k-nearest neighbor classification models, using transcriptomic data across the original cohort and from the Parkinson's Disease Biomarkers Program to predict cluster groupings. Results: The clusters from the analysis displayed robust differences in terms of UPDRS scores, especially regarding motor symptoms. Gene set enrichment and deconvolution analyses revealed differences in immune system activity by cluster. The classification models across cohorts suggested generalizable implications in terms of the UPDRS, supporting the applicability of the results beyond the initial cohort. Conclusions: The correlations between clusters, clinical outcomes, and immune system activity support our hypothesis that molecular data can provide more informative classifications of PD relative to clinical tests alone. To build on these findings, future research can continue to investigate clinical, biological, and practical implications of clusters, potentially with other selections of omics data or in additional cohorts.