Genome based subtyping identifies Parkinson’s disease subtypes with reproducible motor progression differences and clinical trial enrichment potential

Research overview

The text below is the source article’s original abstract under a verified reuse license. cc by

Abstract Parkinson’s disease (PD) is clinically and biologically heterogeneous. We tested whether polygenic variation alone could identify early PD subtypes with distinct longitudinal outcomes and replication cohorts. We analyzed whole genome sequencing from 1,419 participants across the Parkinson’s Progression Markers Initiative (PPMI), Parkinson’s Disease Biomarkers Program (PDBP), and STEADY-PD III trial cohort (STEADY). An autoencoder trained on 1,050 nonanalytic participants learned representations from 98 PD risk and progression variants. Gaussian mixture modeling was then applied to 369 analytic participants, with the PPMI used for discovery (n = 265) and PDBP plus STEADY for validation (n = 104). Three genetically derived clusters were identified (C1, C2 and C3). C1 had the fastest and C3 the slowest MDS-UPDRS III progression in discovery (3.21 vs 2.14 points/year; p = 0.0072), a pattern replicated in validation (3.21 vs 0.49 points/year; p = 0.0034). Gait and posture progression were higher in C1, whereas rigidity progression was higher in C2, across datasets. In extended PPMI analyses, deep brain stimulation was more frequent in C1 than C2 and C3 (21.7%, 14.1%, and 13.6%), with lower age-adjusted hazards in C2 and C3 than C1 (HR 0.40, p = 0.013; HR 0.35, p = 0.0099). Clinical trial simulations suggested that enriching cohorts with C1 participants could reduce required sample size up to 45.4%. These findings support baseline genomic subtyping as a reproducible strategy to identify early PD subgroups with different motor trajectories and treatment-related milestones.

Related topics