Aging/Research reports/2026-10-07
Does the genetic score implicate TIMP2—or other pathways?
Does the genetic score implicate TIMP2—or other pathways?
Today's decision: AG-D06 adds human observational evidence to the TIMP2 candidate. It establishes neither benefit from TIMP2 administration nor selective microglial cargo disposal. AG-C03/AG-T01 remain 🔵 Project Level 2, low confidence. The literature study is ⚪ Level 1; limited human association evidence, no causal efficacy evidence.
One study, different evidence routes
Anastasi et al., Proteomic polygenic risk scores of age-related plasma protein levels reveal a role for Metalloproteinase inhibitor 2 (TIMP2) in cognitive performance, DOI 10.1016/j.neurobiolaging.2025.10.003. First online 23 October 2025; issue January 2026, volume 157, pages 68–78. Retrieved and assessed 7 October 2026. Final original text · Publication and version record. Two linked preprints belong to the same study family, not independent confirmations.
The genetic score weights come from Fenland. Knight-ADRC and ALFA+ validate the prediction of measured proteins. Only ALFA+ supplies the cognitive association here: a higher TIMP2 score is associated with better PACC cognitive composite and episodic memory. Regression models account for age, sex and education. These are cross-sectional associations; they measure neither slower decline nor a treatment effect.
The ALFA+ cohort description includes 410 cognitively unimpaired people enriched for familial Alzheimer's risk. Different measurements and available additional data have different denominators; not every analysis includes 410 people. Predominantly European populations limit generalisability. Two successful protein predictions are not two cognitive replications.
The decisive countercheck
The ALFA+ TIMP2 score contains 119 protein-associated variants: one cis and 118 trans variants, including APOE-rs429358 (original text 3.3–3.4). The score therefore draws mainly on loci outside the immediate TIMP2 gene region. Such variants can capture other relevant pathways. A non-significant single-variant test does not rule out these alternatives. The study reports no Mendelian randomisation.
Measured plasma TIMP2, by contrast, has no significant association with PACC or episodic memory, using either SomaScan or ELISA. This difference does not disprove a TIMP2 effect: genetic regulation, current protein measurements and administered protein are different exposures. It does prevent the simple equation “higher score = effective TIMP2 treatment”. Exact cognitive effect sizes and uncertainty intervals from the supplement have not been verified here; no values were estimated from figures. According to the authors, Aβ-stratified findings do not survive FDR correction.
What changes in the evidence chain?
Our inference: The human association warrants a targeted check, but no upgrade of causality or of AG-T01's mediation hypothesis. Yesterday's mouse findings retain their own limitations; the axolotl linkage gap remains open. A shared mechanism has not been established. Lasting benefit, safety and functional improvement from administration in humans are untested in this study.
Next executable step: Check the existing supplement or available analysis code for analyses excluding the APOE region, using cis variants, changing LD/variant thresholds, or assessing colocalisation. The result should determine how target-specific this evidence route is. An unexamined supplement does not mean those analyses are absent. No experiments or new calculations were performed.
Evidence and limits
- Established within the audit: Score construction and distinct cohort roles are documented in the original text.
- Promising indication: A human score–cognition association supports further scrutiny of regulatory pathways.
- Experimental findings: No new intervention today; earlier animal and cellular findings retain separate assessments.
- Hypothesis: TIMP2 mediates functional benefit; competing genetic pathways remain possible too.
Final main text and metadata assessed, tables read as text; no successful pixel inspection, supplement or raw-data audit. The targeted DOI search found no matching correction or retraction notice, without a completeness guarantee. Source delta +1: 28 selected studies, including one labelled preprint. This is not claimed to be a study published today. Existing professor assignments on human causal inference and evidence dependency cover this refinement; no duplicate package or wish.
Search log · Evidence chain · Approaches registry · Research colloquium.
Continue exploring
