Three large independent snRNA-seq cohorts (SEA-AD ~84 donors DLPFC/MTG, DOI:10.1038/s41593-024-01774-5 confirmed via CrossRef t53a, PMID:39402379 confirmed via EuropePMC t53b; ROSMAP syn3219045; Mathys 2023 GSE174367) each name microglia subtypes using slightly different marker panels and clustering resolutions. Prior literature searches (t53c, t53d) returned zero results with overly specific queries; t54a-t54c broaden to recover Olah 2020 taxonomy baseline and ROSMAP microglia compositional data. TREM2 (target_ref) confirmed as canonical DAM marker via UniProt t54d. Next tick: recover Mathys 2023 GSE174367 dataset metadata and confirm ROSMAP syn3219045 microglia cell count estimates before finalizing kill-criteria thresholds.
Details
- disease
- alzheimers
- target_ref
- gene:TREM2
- identification_strategy
- observational
Raw fields (4)
- assay_spec
planned input: snRNA-seq count matrices from SEA-AD (CellxGene Census, DOI:10.1038/s41593-024-01774-5, crossref status pending t53a), ROSMAP (Synapse syn3219045), Mathys 2023 (GEO GSE174367); planned procedure: subset microglia using P2RY12 (UniProt Q9H244) + CSF1R + CX3CR1 co-expression; Harmony+scVI joint integration across cohorts; Leiden clustering at res 0.3/0.5/0.8; marker dotplots vs DAM (TREM2/SPP1/LPL/GPNMB), homeostatic (P2RY12/TMEM119/CX3CR1), MGnD (APOE/ITGAX/CSF1), IRM (IFIT1/IFI44L), lipid-droplet (PLIN2/PLIN3) signatures from Olah 2020, Sun 2023, Mathys 2023; crosswalk table mapping per-cohort cluster labels to shared MG-1..MG-6 taxonomy; scCODA Dirichlet regression of subtype proportions vs Braak I-VI across all three cohorts jointly; planned output: ARI matrix, UMAP colored by cohort + subtype, composition boxplots by Braak stage, kill-criteria check per protocol
- hypothesis
Named microglia subtypes from SEA-AD (MG-homeostatic, MG-DAM, MG-MGnD, MG-IRM, MG-lipid), ROSMAP snRNA-seq, and Mathys 2023 collapse to a shared taxonomy of 4-6 core states when integrated under Harmony/scVI, and their compositional shifts as a function of Braak stage are reproducible across cohorts after correcting for donor-level variance.
- kill_criteria
Abort or revise if: (1) DOI:10.1038/s41593-024-01774-5 cannot be confirmed as SEA-AD flagship — RESOLVED: confirmed CrossRef t53a + PMID:39402379 EuropePMC t53b; (2) ROSMAP syn3219045 or Mathys 2023 GSE174367 microglia yield fewer than 2,000 post-QC cells (P2RY12+/CSF1R+ gate) — pending dataset access confirmation; (3) Braak stage metadata absent from any cohort — pending; (4) ARI across all three cohorts falls below 0.10 after Harmony+scVI (indicating no shared subtype structure) — to be evaluated post-analysis.
- primary_endpoint
ARI between per-cohort Leiden cluster labels after joint Harmony+scVI embedding (target ARI > 0.25, indicating non-trivial cross-cohort subtype correspondence); Dirichlet composition p-values (FDR < 0.05) per subtype vs Braak stage across >= 2 cohorts; marker gene overlap (Jaccard >= 0.4) between integrated cluster top-20 DEGs and published DAM/MGnD/IRM/homeostatic signatures from Olah 2020, Sun 2023, and Mathys 2023 taxonomies.