Objective: To stratify Parkinson’s disease (PD) patients based on key molecular mechanisms by using a genomic-based approach.

Background: PD is a complex neurodegenerative disease involving several mechanisms. The AETIONOMY project aims to validate the taxonomy of neurodegenerative diseases based on their mechanisms to identify homogeneous groups of PD patients characterized by a combination of candidate mechanisms provided by the consortium.

Method: Clustering analysis was conducted on two independent datasets, the DIGPD cohort (416 PD patients) for hypothesis generation, and the AETIONOMY cohort (350 PD patients) for replication. After genome wide genotyping and imputation, genetic variants were selected according to their functional impact in the brain (CADD score) from the genes involved in five candidate mechanisms (synuclein methylation, mitochondrial dysfunction, neuroinflammation, stress-induced comorbidity, insulin pathway). As a genomic clustering model, we used a derived method from the non-negative matrix factorization (NMF) which is a method of factorization useful for dimension reduction and for clustering simultaneously group of patients and group of variants (mechanisms) in our study. To preserve the constitution of the mechanisms with variant members in the clustering algorithm, we added a prior knowledge through a penalty constraint on the variants.

Results: Analyses were conducted on a total of 901 genetic variants. Using NMF, we determined three clusters of PD patients from DIGPD, which are mainly defined by stress-induced comorbidity and insulin pathway as the main contributors of the clustering. Statistical significant differences were found between clusters for the MDS-UPDRS and frontal assessment battery (FAB) scores. As a replication study using the NMF with the same variants on AETIONOMY, we found similar profile of clustering (3 clusters), contributors (same mechanisms), and clinical differences between clusters

Conclusion: Our genomic-based mechanism-driven stratification of PD identifies 3 clusters of patients with different clinical profiles. The derived mechanism-based stratification is thus expected to provide a valuable guidance for future biological investigations towards a better understanding of the distinct forms of PD.

« Back to 2019 International Congress

MDS Abstracts - https://www.mdsabstracts.org/abstract/mechanism-based-classification-in-pd-cohorts/