A LASSO-optimized composite score integrating plasma GFAP, S100B, NFL, soluble PDGFR-β, and claudin-5 fragments to achieve AUC >0.90 for pre-symptomatic AD/PD, outperforming single-marker approaches including amyloid PET.
Curated pathway from expert analysis
flowchart TD
A["Astrocyte End-Foot GFAP Network<br/>Perivascular Cytoskeleton"]
B["AQP4 Polarization and Claudin-5 Support<br/>BBB Interface Integrity"]
C["Stress-Driven GFAP Redistribution<br/>End-foot to Soma Shift"]
D["Barrier Support Failure<br/>Neurovascular Coupling Weakens"]
E["Plasma GFAP/S100B Rise<br/>Early BBB Dysfunction Readout"]
F["Pre-symptomatic Neurodegeneration Signal<br/>Astrovascular Injury Window"]
A --> B
C -.->|"disrupts"| B
B --> F
C --> D
D --> E
E --> F
style A fill:#1a237e,stroke:#4fc3f7,color:#4fc3f7
style D fill:#b71c1c,stroke:#ef9a9a,color:#ef9a9a
style F fill:#1b5e20,stroke:#81c784,color:#81c784No linked papers recorded for this hypothesis yet.
Median TPM across 13 brain regions for GFAP from GTEx v10.
No clinical trials data linked to this hypothesis yet.
No curated ClinVar variants loaded for this hypothesis.
Run scripts/backfill_clinvar_variants.py to fetch P/LP/VUS variants.
No DepMap CRISPR Chronos data found for GFAP.
Run python3 scripts/backfill_hypothesis_depmap.py to populate.
| Prediction | Predicted | Observed | Status | Conf |
|---|---|---|---|---|
| IF the LASSO-optimized NVPS composite panel (plasma GFAP, S100B, NFL, soluble PDGFR-β, claudin-5 fragments) is applied to a pre-symptomatic AD/PD cohort at baseline, THEN the AUC for predicting conver | AUC >0.90 for NVPS panel in pre-symptomatic AD/PD prediction | — no observation — | pending | 0.45 |
| IF NVPS composite score is directly compared head-to-head against plasma GFAP alone, S100B alone, NFL alone, and amyloid PET SUVr in pre-symptomatic AD/PD participants, THEN NVPS will demonstrate stat | NVPS AUC significantly exceeds AUC of best single marker and amyloid PET | — no observation — | pending | 0.55 |