Description
SEA-AD (Gabitto, Travaglini et al., Nature Neuroscience 2024), ROSMAP (Bennett/De Jager cohort), and Mathys et al. Cell 2023 each report distinct microglia subclusters generated from incompatible integration pipelines, differing Leiden resolutions, and non-harmonized neuropathology staging schemes. The core unresolved question is whether homeostatic, DAM, MGnD, IRM, and lipid-droplet-accumulating subtypes across these three datasets converge to a shared taxonomy when evaluated via marker-gene crosswalks and Braak-stage-stratified compositional trajectories. No published study has performed a joint re-integration (Harmony or scVI) of all three snRNA-seq microglia subsets with systematic marker-gene overlap scoring against independent reference taxonomies (Olah, Sun, Mathys) and Dirichlet compositional regression stratified by Braak and CERAD stage. Until this reconciliation is done, cross-cohort compositional claims — including which subtype expands at late Braak, and whether resilience-associated microglia states are cohort-specific or generalizable — remain unverified.