No published study has directly quantified the temporal coupling between MGnD/DAM microglia expansion and mature oligodendrocyte (MOL) depletion across Braak stages in human snRNA-seq data using within-donor compositional models. This plan formalises the analytical approach needed to fill knowledge gap abca820c-b167-45df-821d-0dad9a87e610. Step 1: Obtain SEA-AD DLPFC snRNA-seq cell-type fraction table (MG subtypes × donor, OL subtypes × donor) annotated with Braak I–VI, CERAD, Thal, age, sex, and PMI from CELLxGENE Census. Step 2: Fit cross-lagged Dirichlet regression (scCODA) of MGnD fraction ~ Braak + covariates and MOL fraction ~ Braak + covariates; test cross-lagged MGnD(Braak t) → MOL(Braak t+1) vs reverse ordering using the ordinal Braak bins as pseudo-time. Step 3: Replicate in ROSMAP snRNA-seq (Synapse syn3219045). Step 4: Stratify by APOE genotype and sex as pre-specified secondary analyses. Step 5: Validate directional ordering using scVI + Milo differential-abundance trajectory on joint MG+OL subset. Green 2024 cellular-community resilience axis serves as independent stratifier for resilience/vulnerability enrichment. Marker gates: MGnD = SPP1+GPNMB+ITGAX+LPL+ (top quartile per donor); MOL = MBP+PLP1+MOG+ high, PDGFRA− (exclude OPC). Confounders: batch (donor cohort), PMI, age, sex, APOE4 dosage. Success criterion: Dirichlet posterior credible interval for MGnD ~ Braak excludes zero at ≥95% in both cohorts; cross-lagged coefficient direction consistent across SEA-AD and ROSMAP.
Details
- disease
- alzheimer's disease
- target_ref
- knowledge_gap:abca820c-b167-45df-821d-0dad9a87e610
- primary_endpoint
- Count table successfully computed from Census obs metadata without downloading X matrix (confirms tool access)
- identification_strategy
- observational
Raw fields (3)
- assay_spec
CELLxGENE Census Python API query: pull obs metadata for SEA-AD human DLPFC cells (Microglia + OPC + Oligodendrocyte supertype labels), count donors per Braak stratum (I–VI), output a donor×stratum×cell-type count table, apply ≥10 donors/stratum gate, publish analysis_result or blocker.; Donor × Braak-stratum × cell-type count table (planned): rows = Braak I/II/III/IV/V/VI, columns = [n_donors_MG, n_cells_MG, n_donors_OPC, n_cells_OPC, n_donors_MOL, n_cells_MOL] for SEA-AD DLPFC; {'kind': 'claim', 'text_template': 'SEA-AD Census slice contains [N] donors across Braak strata; [K] of 6 strata meet the ≥10-donor threshold for both Microglia and Oligodendrocyte fields (planned)'}; analysis_result (planned): 'AD cross-cohort feasibility gate — SEA-AD microglia × OPC stratum counts' — contains the count table, pass/fail per stratum, and a proceed/blocker decision with exact missing-data specification if blocked- hypothesis
Expansion of SPP1+/GPNMB+ disease-associated microglia (MGnD/DAM subtype) precedes or is contemporaneous with depletion of mature oligodendrocytes (MOL; MBP+/PLP1+/MOG+) across Braak I–VI, and this coupling is reproducible across SEA-AD (~84 donors, DLPFC) and ROSMAP snRNA-seq cohorts within a within-donor compositional framework.
- kill_criteria
Abort or revise if: braak_stage field name in Census obs may differ from 'braak_stage' — must inspect obs schema before filtering; fallback to 'Braak_stage' or free-text disease_stage field; SEA-AD DLPFC may not be the tissue label used in Census — may need 'middle temporal gyrus' or wildcard tissue match; ROSMAP Braak metadata may not be present in Census obs for the ROSMAP slice — if absent, gate applies to SEA-AD only and ROSMAP feasibility is marked 'requires Synapse access'; Census version may not include latest SEA-AD release — record census version string in the output artifact