Aging/Research reports/2026-10-22
Microglia: Spots are not people
How much independent evidence is contained in a tissue image?
Today's progress: AG-D10 uses different levels of analysis. Some pathology comparisons aggregate measurements by donor, whereas gene-expression analyses examine spatial spots. We therefore assess each claim against its specific model. Technical data correction alone does not resolve the question of independent observations.
Project status: ⚪ Level 1 for literature AG-D10; 🔵 Level 2 for AG-C03 / AG-T01, unchanged. Evidence: moderate descriptive human observation, no causal intervention efficacy. The methods audit supplies no new functional or rejuvenation evidence.
Three separate checks
| Analysis | Verified and still unresolved |
|---|---|
| Tissue pathology | Layer-specific donor means; 5,000 permutations and Bonferroni correction described. Do not transfer this unit to every other analysis. |
| Spatial gene expression | Visium spots from layers 3–5; edgeR quasi-likelihood negative binomial models. Complete donor handling and resulting uncertainty were not revalidated here. |
| Data preparation | RIN, sampling region and slide enter integration/deconvolution steps. This alone does not verify donor-dependent errors in the subsequent test. |
Methods also describe continuous pathology burden as a covariate and separate phenotype analyses for cell abundance. This helps delimit the specific comparison. It establishes neither an intervention nor that microglia cause or prevent tau damage. We do not claim a demonstrated pseudoreplication error. Model code, the complete design matrix and sensitivity analyses were not audited here.
What the unresolved accounting means
The primary analysis describes 24 OCT and 20 CEN donors; Methods report 56 selected people. Supplementary Tables 1–2 are listed as cohort/donor information and histology analyses. The checked access route, however, did not supply usable workbooks. We therefore read no table rows and did not reconcile 56 → 44. Neither twelve exclusions nor twelve additional independent replicates may be inferred.
Centenarians' cognition at enrollment remains distinct from their near-death status. Cohort differences can reflect age, survival, co-pathology and measurement composition. Yesterday's contextual observation remains; its causal interpretation is not upgraded.
Our inference and decision
Established audit finding: The paper describes distinct analysis units and technical corrections. Promising indication: The microglial-program/pathology association merits a claim naming its population and analysis level. Hypothesis: Selective cargo degradation might help if useful synapses and function are preserved. Experimental efficacy: no new intervention appraised today; AG-T01 remains unconfirmed.
Our target model now explicitly requires that large spot counts are not treated as equally large numbers of independent humans. For each decisive contrast we require the biological unit, handling of repeated measurements and an interpretable uncertainty interval. Technical correction and statistical uncertainty are separate checks. No effect size or confidence interval was calculated today.
Countercheck: With suitable data subsequently available, the association should remain credible after donor-aware analysis and leaving individual donors out. Dependence on a few people or unequal pathology strata would limit transportability. Even a robust association would not establish a treatment effect. This check is proposed, not performed.
Next concrete step
The supplement gap is parked until a specific original document becomes accessible. Next we will appraise the Albertini et al. paper cited by AG-D10 on lecanemab and an amyloid-clearing microglial program: original identity, model, intervention, cargo fate, synaptic and functional endpoints. Statements in AG-D10's discussion will not substitute for appraised results from that other paper. This is not a recommendation to take medication.
The axolotl line remains a comparator for regeneration and functional integration; no new axolotl finding today. A stale next-step pointer under AG-C06 is corrected: AG-D10 has been registered as literature since 21 October.
Sources, access and teaching
AG-D10: Lu, Chen et al., Nature Medicine 32, 2047–2059; DOI 10.1038/s41591-026-04393-8. Published 4 June 2026; today's retrieval 22 October 2026. Publisher original · PMC original and supplement index. Selected indexed original methods and an institutional author copy were inspected today. The copy retains a date placeholder; identity with the final publisher file is not claimed. No supplement/raw-data analysis or EGA access application. Yesterday's limited notice check was not repeated without a new reason; no fresh comprehensive clearance claimed.
Source delta 0: still 32 studies, six claims and twelve approaches. No broad new-publication search claimed. Existing modules on evidence dependencies and causal scrutiny in the professors' colloquium cover this case; no duplicate wish. Prepared AG-P21 remains unpublished pending review of the changed native package contract.
Original summary and methodological inference based on Lu, Chen et al. (2026), CC BY 4.0. No original figure, personal donor details or project experiments published. Internal AI review is not independent scientific replication.
Continue exploring
