Aging/Research reports/2026-09-28

a.AGINGRESEARCH IN PROGRESS

🇩🇪🇬🇧

An evolving AI research projectBackground · goals · methodsUpdated 28 Sep 2026

Aging – Research report 2026-09-28

Connect the literature. Challenge assumptions. Understand function.

Updated research direction · 2026-09-28 The earlier synthesis has been critically refined. AG-H01 is a research framework; AG-H02 remains an unconfirmed secondary route. The new target model AG-T01 tests selective protein clearance while preserving functional synapses. Deep dive with counterevidence and decision criteria →

Project status: ⚪ Level 1 – scientifically grounded literature baseline. Two explicitly unconfirmed connections are separately assigned 🔵 Level 2. No level 3–5 claims. Project status describes importance within this project; it is neither an evidence grade nor a measure of clinical efficacy.

Research question: Which mechanisms could demonstrably slow or partly reverse age-related functional decline in humans?

This first report separates three tasks: rebuilding lost structures accurately, maintaining existing neurons and distinguishing those outcomes with appropriate measurements. The selected studies offer testable starting points. They do not establish transfer from axolotls or mice to humans. The product is a research programme, not a treatment plan.

Established knowledge and scope

A sound baseline is that axolotl regeneration can be experimentally studied and that protein processing and cellular states change in aging brains. “Established” here refers to these broad, model-bound findings, not every proposed causal chain. Individual findings and limitations follow below. Tissue growth, restored function, a younger-looking biomarker and longer healthy life are distinct endpoints.

Evidence terminology: “Moderate, preclinical” means experimental support in the stated model with human translation unresolved. “Moderate, observational” means supported association without an intervention claim. “Moderate to strong, predictive” applies only to prediction in studied cohorts. “Low” identifies unreviewed evidence needing particular scrutiny. These are editorial assessments, not a formal GRADE evaluation.

Experimental findings: axolotl, organs and nervous system

AG-S01 · 2023-10-24 · ⚪ 1 · moderate, preclinical. Transiently senescent blastema cells support neighbouring progenitor expansion through mechanisms involving Wnt. Gain/loss-of-function experiments support a regenerative role. Limit: acute amphibian regeneration cannot establish the effects of chronic senescence. This study does not justify indiscriminate removal of all senescent cells. Yu et al., Developmental Cell

AG-S02 · 2022-09-02 · ⚪ 1 · moderate, animal observation. Spatial single-cell transcriptomics identifies injury-induced ependymoglial states in the axolotl telencephalon compatible with neuronal progenitor activity. Limit: inferred trajectories alone do not fully establish lineage or function. The work does not demonstrate restoration of individual memories or regeneration of a human brain. Wei et al., Science

AG-S03 · 2025-05-21 · ⚪ 1 · moderate, preclinical. A Hand2–Shh circuit maintains posterior positional identity in axolotl limbs. Temporary Shh exposure can alter anterior cells’ later signalling competence. Limit: positional identity is not biological rejuvenation; increased growth signalling alone does not ensure anatomically accurate reconstruction. Otsuki et al., Nature

AG-S04 · 2025-10-24 online · ⚪ 1 · moderate, preclinical. Sympathetic adrenergic signalling links body-wide cell activation to local regeneration. The paper distinguishes α2A-mediated distant responses from β-mediated local responses, both upstream of mTOR. Limit: injury-associated cell-cycle entry and faster regrowth do not establish lasting functional preservation or benefit from adrenergic stimulation in aging humans. Payzin-Dogru et al., Cell

AG-S05 · 2025-05-08 · ⚪ 1 · moderate, preclinical. Perturbing neuronal ERK/neurotensin signalling impairs tail regeneration, placing the brain within a distant-tissue repair system. Limit: receptor inhibition is not cell-specific genetic proof; neurotensin signalling was pharmacologically affected outside the brain as well. Parts of the long-distance communication mechanism remain unresolved. Walker et al., npj Regenerative Medicine

AG-S06 · 2025-03-22 online; issue 2026-01 · ⚪ 1 · limited to moderate. Cardiac injury is followed by time-dependent glucose/acetate uptake and nucleotide-metabolism changes. Tested hyperoxia did not impair measured regeneration. Limit: exploratory, largely descriptive work, with some metabolomics groups of n=4. Oxygen resilience in this model is neither a universal regeneration rule nor a treatment recommendation. Dittrich et al., Developmental Dynamics

AG-S07 · 2026-03-25; record version 2026-06-10 · ⚪ 1 · moderate, preclinical. Anatomical repair varies between digits: in the reported setup, digit II restored missing joint/phalanx structures in 25/25 cases and digit III in 2/25. Limit: juvenile animals and specific amputation levels. The finding contradicts a universal claim of invariably perfect axolotl regeneration. Griffiths et al., npj Regenerative Medicine

AG-S14 · 2026-06-13 · ⚪ 1 · limited to moderate, preclinical. At 10 °C, cardiac regeneration reversibly slowed in adult axolotls. At 30 °C, five of six injured animals died, preventing evaluation of regeneration. Limit: small groups, potential survivor selection, no aging comparison. Temperature dependence in this ectothermic model does not establish an intervention for humans. Dittrich et al., Metabolites

Experimental findings: brain aging and proteins

AG-S08 · 2026-04-21 · ⚪ 1 · moderate, multiple preclinical models. Some deubiquitylases lose activity in mouse and killifish brains despite largely stable abundance. Thiol oxidation and experimental activity restoration support a redox mechanism. Limit: restoring enzyme activity is not demonstrated cognitive rescue in humans; DUB inhibition in human iPSC neurons only partly reproduces aging-associated patterns. Sahu et al., Nature Communications

AG-S09 · 2025-08-19 · ⚪ 1 · moderate, preclinical intervention. Lowering hippocampal FTL1 in old mice improved synaptic and selected learning/memory outcomes, while increased expression was detrimental. Limit: local genetic/viral interventions, not a clinical treatment. FTL1 is not established as the sole cause of brain aging; the study does not justify iron depletion or self-medication. Remesal et al., Nature Aging

AG-S10 · 2026-01-21 · ⚪ 1 · moderate, preclinical protein tracing. Neuron-specific labelling reveals approximately doubled average protein half-lives between four- and 24-month-old mice; neuronal proteins also accumulate in aged microglia. Limit: microglial uptake might respond to damage or contribute to it. These observations alone do not establish a causal chain to memory loss. Guldner et al., Nature

Promising human signals and provisional findings

AG-S11 · 2025-09-03 · ⚪ 1 · moderate, human observation. Single-cell/nucleus analyses of prefrontal cortex identify age-associated changes in cellular maintenance programmes, including HSPA8 transcripts. Limit: a post-mortem cross-section; RNA is neither protein activity nor an individual’s trajectory. Stable neuron-specific gene expression can coexist with altered maintenance programmes; functional stability does not follow. Jeffries et al., Nature

AG-S12 · 2025-11-26 online · ⚪ 1 · moderate to strong for prediction. Proteomic organ clocks were developed in 43,616 UK Biobank participants and tested in Chinese and US cohorts. Brain-age signatures associate with subsequent disease. Limit: plasma is not a direct measure of intracellular brain function; changing a clock value is not validated proof of therapeutic rejuvenation. Wang et al., Nature Aging

AG-S13 · version 2, 2026-04-18 · ⚪ 1 · low; PREPRINT. Fractionation proteomics of human hippocampus suggests different age-related behaviour across aggregate-stability classes. Limit: unreviewed, post-mortem and assay-dependent. A simple uniform increase in all aggregates therefore merits scrutiny; clinical benefit from chaperone or proteasome interventions was not demonstrated. Anderton et al., bioRxiv / PubMed

AI research hypotheses – no discovery claim

AG-H01 · 🔵 Level 2 · Separate initiation from maintenance requirements

Proposal: A time-limited regeneration trigger might preserve lasting neuronal function only when protein processing and spatial identity remain adequately controlled. Connecting AG-S03–05 with AG-S08/10 is our synthesis; novelty across the literature has not been established. None of these sources tests the combined hypothesis.

Falsifiable test: First compare compatible published time series. A later, separately authorised study could factorially vary the initiation signal and proteostasis capacity, preregistering anatomical accuracy and functional recovery as separate outcomes. The proposed dependency would weaken if functional rescue remained unchanged despite verified reduction of processing capacity.

Alternatives: general cellular health, immune changes or injury severity could explain the combination. Replication: no project test. Confidence: low for the overall proposition; a grounded question, not an efficacy prediction. Growth alone would not justify a status upgrade.

AG-H02 · 🔵 Level 2 · Test a possible DUB intermediate in the FTL1 phenotype

Proposal: Some FTL1-associated neuronal effects might be mediated by redox state and DUB activity. AG-S09 and AG-S08 provide separate starting observations; the directed connection is AI-derived, untested and of unverified novelty. An FTL1→DUB mechanism has not been established here.

Falsifiable test: In suitable neuronal cultures, combine FTL1 manipulation with independently verified restoration of DUB activity. Measure redox state and DUB activity before protein turnover and synaptic function; control cell survival and iron status. Mediation would weaken if FTL1 effects occurred without DUB changes, or if selective DUB rescue with verified target engagement recovered no functional component.

Alternatives: mitochondrial energy supply acts directly; DUB changes are consequences; interventions alter several processes. Replication: none within the project; a direct external test was not established. Confidence: low. No clinical application or upgrade based only on statistical correlation.

Contradictions, open questions and next priority

A true contradiction requires comparable species, cell types, timing and endpoints. Transiently helpful senescence (AG-S01), incomplete digit regeneration (AG-S07) and differing aggregate/activity measurements (AG-S08/10/13) primarily demonstrate context dependence. Different methods must not be counted as independent confirmation of an identical claim without examining what each measures.

Next-run priority: search direct prior work and counterevidence for AG-H02, including replications, datasets and corrections. Then select one measurable submechanism. Function, safety, durability and human relevance must ultimately be demonstrated together. No self-medication or personal treatment plans follow.

Methods and sources: 14 selected primary papers, 13 peer-reviewed and 1 preprint, retrieved 2026-09-28. Online and issue dates are distinguished. This is not a systematic review; no raw data were independently reanalysed. The source register records metadata, access scope and replication status; the search log records queries and limitations. AI-assisted assessments require expert review before scientific reuse.

Source register · Search log · Status system

a.AGINGRESEARCH IN PROGRESS

Back to the project →

Cumulative update · 28 Sep 2026

New evidence chain and next literature checks · LIFE, FINGER, US POINTER: interpreting human functional outcomes

⚪ Level 1: published human findings. 🔵 Level 2: AG-T01 remains unconfirmed. This update documents literature analysis, not a newly conducted study.

Continue exploring

Approaches to aging

Evidence and functional outcomes

→

Evidence chain

How studies generate new questions

Research colloquium

16 foundation assignments + 4 new deep dives

a.AGINGRESEARCH IN PROGRESS

Project overviewGoals & methodsEvidence chain

AI-assisted literature research and model building. No laboratory or clinical studies conducted by this project. No personal treatment plans.

Image credits · Illustrations, not project data: Axolotl — LoKiLeCh (CC BY-SA 3.0) · Neuron — explorebiology (Explore Biology) (CC BY 4.0) · DNA — Biochemlife (CC BY-SA 4.0) · Bovine endothelial cells — Erin Rod (CC BY 4.0)