Aging/Research reports/2026-10-11
TIMP2: What two datasets cannot establish together
Project status: 🔵 Level 2 · AG-C03 / AG-T01. Our synthesis requires further testing. Scientific evidence: preclinical literature; today a bounded metadata audit. No new efficacy finding or independently confirmed project mechanism.
The central finding
Two publicly deposited datasets from the TIMP2 microglia paper address different questions. Connecting them may suggest hypotheses; it cannot identify a joint genotype-by-treatment effect. Today’s advance is a concrete boundary for the next data analysis.
Two series, different contrasts
| Data source | Documented comparison and limit |
|---|---|
| GSE297916 | Hippocampus, 20-month-old male wild-type mice, TIMP2 versus PBS, single-nucleus RNA sequencing. Four deposited libraries, two per arm. |
| GSE297917 | Primary microglia, postnatal days 6–7, wild type versus TIMP2 knockout, bulk RNA sequencing. Fourteen pooled samples, seven per genotype, each from five pups. |
| Our inference | Age, material and intervention differ. Both series belong to one paper; they constitute neither a joint factorial experiment nor two independent studies. |
Experimental units before significance
The directly read snRNA examples GSM9001952 and GSM9001953 describe pooling two biological samples receiving the same treatment after nuclear isolation: 10,000 loaded nuclei from each. These are neither 20,000 independent treated animals nor necessarily 20,000 finally analysed nuclei. The second replicate pair has not been fully checked; no total number of distinct mice is inferred.
The bulk example GSM9001960 describes pooled male and female neonatal microglia. Five pooled animals yield one combined expression profile, not five separately observed profiles. Individual mouse/litter identifiers, pool membership and complete sex/batch balance are unverified.
How the conclusion changes
Our inference: A difference between these series isolates neither an age effect nor therapeutic reversal of a knockout effect. Alternatives include developmental stage, cell composition, culture and technical processing. Pooling may hide additional dependencies. This does not refute the existing external findings; it narrows which comparisons are justified.
Established knowledge for this audit: the specified metadata fields. Promising indications: earlier TIMP2 functional findings with their limitations. Experimental results: external animal/cell experiments, with no newly analysed measurements today. Hypothesis: selective harmful-cargo degradation with preserved synapses and improved function. AG-T01 remains at Level 2 with low mechanistic confidence. Persistent benefit after withdrawal, human efficacy and adequate safety are not established.
The next executable check
Check GSM9001954/55 and a WT bulk comparator; tabulate group, age, sex, batch, pool members/mouse ID, library and file. Missing identities remain unknown. Only then assess a bounded within-series analysis. Pseudobulk is an option only for suitable cell/nucleus data. Two deposited snRNA samples per arm constrain uncertainty estimation. Do not construct a joint interaction analysis or pooled effect from the different groups.
Source status and research colloquium
- Original paper AG-R01: Hemmer, Philippi et al., Nature Communications, published 12 August 2026; DOI 10.1038/s41467-026-74906-z.
- Dataset dates: both GEO series submitted 22 May 2025, public 1 June 2026, last updated 2 September 2026. These dates are neither article publication nor evidence of new experiments.
- Retrieved/audited: 11 October 2026. Indexed official series/GSM texts; direct access limited by reCAPTCHA/IDP. No matrix/raw-data analysis or complete new methods appraisal. The correction/retraction check of 10 October was not repeated. Source delta 0; 28 distinct studies. Search limitations and unappraised axolotl lead.
- New teaching draft AG-P21: “Aging – Pooling, pseudobulk and non-identifiable interactions”. Prepared in both languages; supplement package and wishlist entry are not yet published. Existing professor assignments remain available. Evidence chain · Approaches registry.
Continue exploring
