Parkinson's Disease Subtype Classification — Precision Medicine Approach

Hypothesis

[Parkinson's disease](/diseases/parkinsons-disease) is not a single homogeneous disorder but comprises distinct biological subtypes with different underlying mechanisms, disease trajectories, and treatment responses. Defining these subtypes will enable precision medicine approaches.

Gap Addressed

PD Cure Roadmap Gap #5 (28 pts): Is PD one disease or several distinct syndromes?

Rationale

Clinical heterogeneity in [PD](/diseases/parkinsons-disease) has long been recognized, but biological basis for subtypes remains unclear:

Hypothesis

[Parkinson's disease](/diseases/parkinsons-disease) is not a single homogeneous disorder but comprises distinct biological subtypes with different underlying mechanisms, disease trajectories, and treatment responses. Defining these subtypes will enable precision medicine approaches.

Gap Addressed

PD Cure Roadmap Gap #5 (28 pts): Is PD one disease or several distinct syndromes?

Rationale

Clinical heterogeneity in [PD](/diseases/parkinsons-disease) has long been recognized, but biological basis for subtypes remains unclear:

Experimental Design

Aim 1: Multi-Modal Dataset Integration

Approach: Create comprehensive dataset for subtype discovery

Cohort: 2000 newly diagnosed PD patients (within 2 years), treatment-naive

Data Collection:

| Modality | Samples | Analysis |
|----------|---------|----------|
| Genetics | Blood | Whole genome sequencing, polygenic risk score |
| Clinical | All | MDS-UPDRS, non-motor scales, neuropsychology |
| Imaging | [MRI](/diagnostics/magnetic-resonance-imaging), DaTscan, [DAT](/proteins/dopamine-transporter), MRI | Structural, functional, diffusion |
| CSF | 1000 patients | [Alpha-synuclein](/proteins/alpha-synuclein), [tau](/proteins/tau), [NfL](/biomarkers/neurofilament-light-chain-nfl), cytokines |
| Blood | All | Multi-omics (proteomics, metabolomics) |
| Digital | Subset (n=500) | Wearable sensors, smartphone tests |

Data Integration:

• Machine learning clustering (UMAP, PHATE)
• Multi-modal variational autoencoders
• Consensus clustering across modalities

Aim 2: Biological Subtype Validation

Approach: Validate subtypes with mechanistic studies

Methods:

• Test drug responses in dopaminergic neurons
• Characterize alpha-synuclein handling
• Measure mitochondrial function

• 100 brains from well-characterized PD patients
• Map alpha-synuclein, tau, neuroinflammation patterns
• Correlate with clinical subtype

Aim 3: Subtype-Specific Treatment Matching

Approach: Test whether subtype assignment predicts treatment response

Retrospective Analysis:

• Analyze clinical trial data by subtype
• Did specific subtypes respond better to specific treatments?
• Example: LRRK2 carriers → LRRK2 inhibitors; GBA carriers → chaperones

• Assign patients to treatments based on subtype
• Compare outcomes to standard of care
• Primary: clinical progression at 2 years

Aim 4: Subtype-Specific Biomarker Development

Approach: Develop blood-based tests for subtype classification

Approach:

• Train classifier on blood biomarkers to predict imaging/CSF subtype
• Validate in independent cohort
• Target: >90% accuracy for major subtypes

• Inflammatory cytokines (IL-6, TNF-α, IL-1β)
• Neurodegeneration markers (NfL, p-tau181)
• Metabolic signatures (lipids, amino acids)
• Genetic risk scores

Subtype Hypotheses

Based on existing literature, expect 4-6 major subtypes:

| Subtype | Key Features | Approximate % |
|---------|--------------|---------------|
| LRRK2-like | Slow progression, typical tremor-dominant, less cognitive decline | 15-20% |
| GBA-like | Rapid progression, early autonomic dysfunction, high dementia risk | 10-15% |
| SNCA/MAPT | Early postural instability, high tau co-pathology | 10-15% |
| Mixed/Other | Does not fit clear genetic pattern | 50-60% |

Key discriminator variables:

• Age at onset
• Motor phenotype (tremor-dominant vs PIGD)
• Non-motor features (cognitive, autonomic, sleep)
• Imaging patterns
• Biomarkers

Expected Outcomes

Scoring

| Dimension | Score | Rationale |
|-----------|-------|------------|
| Mechanistic Impact | 8 | Would transform understanding of PD heterogeneity |
| Cure Proximity | 8 | Enables precision medicine for PD |
| Feasibility | 8 | Large cohort feasible with existing infrastructure |
| Cost Efficiency | 7 | Expensive but transformative for field |
| Timeline | 5 | Full validation 5-7 years |
| Cross-Disease Value | 7 | Relevant to AD, ALS, FTD (similar heterogeneity) |
| Biomarker Enablement | 9 | Strong blood-based biomarker potential |
| Combinability | 8 | Complements genetic and biomarker studies |
| De-risking Value | 9 | Subtype-specific trials reduce failure risk |
| Novelty | 8 | First comprehensive multi-modal subtype analysis |

Total: 77/100

Risks and Mitigations

| Risk | Mitigation |
|------|------------|
| Subtype instability over time | Longitudinal follow-up, assess subtype drift |
| Overfitting in clustering | Validation in independent cohorts |
| Biological sample quality | Standardized protocols, central lab |
| Clinical heterogeneity | Large sample size, diverse sites |

Cost Estimate

| Component | Cost (USD) |
|-----------|------------|
| Cohort recruitment (2000 PD + 500 controls) | $1.5M |
| Whole genome sequencing | $1M |
| Imaging (MRI, DAT scan) | $1.5M |
| CSF analysis (1000 samples) | $750K |
| Multi-omics (proteomics, metabolomics) | $1M |
| iPSC derivation and differentiation | $1.5M |
| Post-mortem brain collection | $500K |
| Personnel (5 FTE) | $2M |
| Data analysis and ML | $750K |
| Total | $10M |

References