Plan: (1) Extract microglia nuclei from CELLxGENE Census slices for SEA-AD and Mathys 2023, plus ROSMAP snRNA-seq (Synapse syn3219045). (2) Harmonize metadata to the donor-spine schema (Braak I-VI, CERAD, Thal, sex, APOE genotype). (3) Integrate with Harmony; cluster at Leiden resolutions 0.3/0.6/1.0; compute silhouette and ARI-to-published-reference. (4) Cross-walk cluster marker genes against Olah 9-state, Mathys homeostatic/DAM, and SEA-AD MG-1–MG-9 panels. (5) Run scCODA / Dirichlet regression: subtype proportion ~ Braak + sex + APOE + (1|donor). (6) Report compositional p-values, effect sizes, and cross-cohort concordance for each subtype. Success criterion: ≥3 subtypes replicate across all three cohorts with marker-gene Jaccard ≥0.4 and composition trend direction concordant (p<0.05) in ≥2/3 cohorts.
Details
- disease
- alzheimer's disease
- identification_strategy
- observational
Raw fields (4)
- assay_spec
Tick 8 — literature crosswalk seeding. Crossref DOI lookups for Olah et al. 2020 (doi:10.1038/s41593-020-0535-5, nine-subtype human microglia taxonomy) and Gabitto/Travaglini et al. 2024 SEA-AD (doi:10.1038/s41593-024-01774-5) issued to confirm canonical references. PubMed searched for (a) Olah 2020 microglia subtypes and (b) DAM/MGnD/IRM/homeostatic crosswalk papers 2023–2025. Results will populate the subtype crosswalk table seeded in Tick 7. Next tick: create notebook artifact with crosswalk skeleton mapping Olah MG1–MG9, Mathys homeostatic/DAM subtypes, and SEA-AD microglia clusters to unified MG-1…MG-N labels using marker genes P2RY12, TMEM119, TREM2, SPP1, LPL, GPNMB, ITGAX, CX3CR1, CSF1R as anchors.
- hypothesis
Named microglia subtypes identified independently in SEA-AD (Gabitto et al. 2024 Nature Neuroscience), ROSMAP (Mathys et al. 2019, 2023), and Olah et al. 2020 collapse onto a shared low-dimensional taxonomy when integrated with Harmony or scVI, and compositional shifts of disease-associated microglia (DAM/MGnD/TREM2-high) states scale monotonically with Braak stage across all three cohorts after controlling for donor-level random effects.
- kill_criteria
Abort or revise if: Substrate artifacts may contain schema descriptions rather than row-level donor data — if so, the verdict is BLOCKED with note: 'schema-only artifact, no per-donor counts available'; Braak staging field name may differ across crosswalk versions (braak_stage vs braak_score vs neuropath_braak); record exact field names found; OPC / remyelination subtype labels may be absent from crosswalk if it covers microglia only — flag as a separate blocker
- primary_endpoint
Tick 8 success criteria: (a) Crossref returns valid metadata for both canonical DOIs confirming publication year, journal, and author lists. (b) PubMed returns ≥2 papers covering cross-cohort microglia subtype naming or DAM/MGnD/IRM taxonomy in human snRNA-seq, adding to the ≥3 already seeded in Tick 7. (c) plan assay_spec updated to reflect accumulated literature scaffold.