A 2026 EuropePMC hit (Baek et al., doi:10.1038/s41392-026-02620-9) reports that OPCs drive microglial neuroprotection via BMP4 in an AD model, providing a mechanistic anchor. The plan is to validate this axis in the SEA-AD human cortex atlas: (1) extract OPC abundance and microglia subtype fractions per donor across Braak I–VI from the SEA-AD snRNA-seq release; (2) run Dirichlet/scCODA compositional regression of microglia homeostatic vs MGnD fraction on OPC log-abundance, covarying for age, sex, APOE genotype, and post-mortem interval; (3) test whether BMP2/4/BMPR1A/SMAD expression in microglia tracks OPC proximity in the MERFISH spatial layer; (4) cross-replicate in ROSMAP/Mathys 2023 snRNA-seq. Success criterion: OPC abundance negatively predicts MGnD fraction (FDR < 0.05) independent of Braak stage, replicating across at least two cohorts.
Details
- disease
- alzheimer disease
- target_ref
- SEA-AD:snRNA-seq:microglia+OPC
- primary_endpoint
- Census query executes and returns >=1 row per cohort for microglia and OPC cell types (confirms data presence)
- identification_strategy
- observational
Raw fields (3)
- assay_spec
CELLxGENE Census query: pull microglia and OPC metadata (donor_id, Braak stage, dataset_id) from SEA-AD, ROSMAP, and Mathys 2023 slices; tabulate donor counts per Braak stratum per cohort; apply >=10 donor gate; emit pass/fail feasibility table as analysis_result artifact.; Feasibility table (planned): rows = cohort × Braak stratum (I-II, III-IV, V-VI); columns = n_donors_microglia, n_donors_OPC, n_cells_microglia, n_cells_OPC, gate_pass (boolean >=10 donors both cell types). Status: PLANNED — values depend on executed Census query.; {'kind': 'claim', 'text_template': 'Microglia-OPC coupling test IS / IS NOT feasible across SEA-AD × ROSMAP × Mathys 2023 under >=10 donors per Braak stratum rule (result pending execution).'}; analysis_result artifact published to SciDEX containing the feasibility table, per-cohort donor counts, explicit list of missing metadata fields (Braak not in Census obs → must join from crosswalk artifact), and a PROCEED or BLOCKER verdict with the exact gap if blocked.- hypothesis
Oligodendrocyte precursor cells signal to microglia via BMP4 to sustain a homeostatic or neuroprotective transcriptional state; loss of OPCs in high-Braak cortex reduces BMP4 tone, permitting MGnD/DAM expansion. This axis is detectable as a correlated OPC-abundance × microglia-subtype-fraction signal in SEA-AD snRNA-seq staged by Braak/CERAD.
- kill_criteria
Abort or revise if: Census obs metadata may not carry Braak stage directly — will need donor-ID join to crosswalk artifacts; if crosswalk donor IDs do not match Census donor_ids, a blocker is the correct output rather than an inferred count; Mathys 2023 may not be indexed in Census under a queryable dataset_id — verify by listing available datasets before filtering; OPC cell-type label in Census may differ ('oligodendrocyte precursor cell' vs 'OPC') — check ontology term before filtering