Tick 25: Hypothesis artifact published (MGnD expansion precedes OPC depletion across Braak trajectory). All three anchors resolved: (1) SEA-AD flagship CONFIRMED, (2) Braak-staged microglia CONFIRMED (Wachter 2024), (3) OPC arm CONFIRMED-COGNITIVE-STATUS (Mathys 2023 Cell, 471 citations) + converging (Green 2024 Nature). Hypothesis linked to this plan. Next milestone: CELLxGENE Census metadata inspection to confirm Braak_stage obs column availability in SEA-AD slice before programmatic composition query.
Details
- disease
- alzheimer's disease
- target_ref
- dataset:SEA-AD-snRNA-seq
- identification_strategy
- observational
Raw fields (4)
- assay_spec
CELLxGENE Census programmatic query (planned): filter by dataset_title containing 'SEA-AD' and 'Mathys'; restrict cell_type to CL:0000129 (microglial cell), CL:0000128 (oligodendrocyte), CL:0002453 (OPC); pull obs columns [donor_id, Braak_stage, cell_type, dataset_id]; count donors per Braak stratum per cell class; emit pass/fail table against kill criteria (>=10 donors per stratum, non-null Braak field). ROSMAP arm: attempt same filter with dataset_title containing 'ROSMAP'; if absent, record NOT_IN_CENSUS and file cohort-specific blocker. Planned output: donor-by-Braak count matrix per cell class; proportional composition table per donor; Spearman rho (MGnD fraction vs Braak) with 95% CI via bootstrap; Dirichlet regression p-value for cell-class ~ Braak as ordinal predictor. Success criterion: SEA-AD microglia arm returns >0 rows with non-null Braak field and >=2 donors per stratum for Braak III-V.
- hypothesis
In the SEA-AD snRNA-seq corpus, MGnD/DAM-state microglia (TREM2-hi, SPP1+, LPL+) expand in proportion earlier in the Braak trajectory (stages III–IV) than the detectable depletion of OPC and mature oligodendrocyte populations, implying a primary neuroinflammatory trigger rather than a secondary response to myelin debris. Cross-cohort replication in ROSMAP and Mathys 2023 (GSE174367) would test whether this ordering is reproducible.
- kill_criteria
Abort or revise if: (1) CELLxGENE Census SEA-AD obs lacks Braak_stage or equivalent harmonized staging column — would require direct h5ad fetch from SEA-AD portal and manual annotation join. (2) MGnD marker genes (TREM2, SPP1, LPL, GPNMB) absent from Census var index for SEA-AD slice. (3) Donor n per Braak stratum <10 in Census SEA-AD slice — underpowered for Dirichlet regression; would require pooling with Mathys 2023 cohort.
- primary_endpoint
ANCHOR STATUS — (1) SEA-AD flagship: CONFIRMED (Gabitto/Travaglini Nat Neurosci 2024, DOI 10.1038/s41593-024-01774-5). (2) Microglia Braak-staged: CONFIRMED (Wachter et al., Acta Neuropathol 2024, DOI 10.1007/s00401-024-02704-2). (3) OPC/oligo composition: CONFIRMED-COGNITIVE-STATUS (Mathys 2023 Cell, DOI 10.1016/j.cell.2023.08.039, 471 citations) + GREEN-2024-CONVERGING (DOI 10.1038/s41586-024-07871-6). HYPOTHESIS ARTIFACT PUBLISHED tick 25. Next: Census metadata inspection for Braak_stage column, then scCODA composition analysis.