Geroscience Repurposing: FDA-Approved Drugs Against the Hallmarks of Aging

Overview

Geroscience holds that biological aging is the dominant modifiable risk factor for the major chronic diseases, and that targeting it should delay several of them at once. A drug that does this is a gerotherapeutic. The practical question is which of the drugs we already have is the best candidate to test, and that question has an answer with a method behind it: Kulkarni, Aleksic, Berger, Sierra, Kuchel and Barzilai, Geroscience-guided repurposing of FDA-approved drugs to target aging (Aging Cell 2022, PMID:35343051), screened DrugAge for compounds that significantly extend rodent lifespan, kept only the FDA-approved ones, and scored the survivors on a 12-point scale.

This project brings that result into dismech. The KB already has the hallmarks of aging as mechanism modules under kb/modules/. What it largely lacked was the drug side: as of the start of this project, 9 of 13 aging-related modules carried no treatments at all, so a query like "what pharmacologic interventions act on inflammaging?" returned nothing, even though the answer is well documented.

The unit of work here is a treatment with target_mechanisms on an aging module — a specific drug asserted to act on a specific hallmark node, with an effect direction and cited, snippet-verified evidence. That turns a review table's tick mark into something the pathograph can draw and a query can reach.

Why the review's own scoring is recorded but not curated as evidence

The 12-point score is a prioritization, not a mechanistic claim, and it is a snapshot: the authors themselves note that "future studies may change the priority order for drugs that did not receive points due to the paucity of clinical data." Several drugs score 0 on human endpoints because nobody has run the trial at the right dose in the right population — not because the trial was run and was negative. Aspirin is the clearest case: the animal work uses anti-inflammatory doses, while every large human trial used antiplatelet doses, which the authors argue is the wrong comparison rather than a refutation.

So the ranking lives in this project file and in module treatment description prose. Evidence items cite the primary studies, which is where the mechanism claims actually are. Citing the review for a mechanism it merely tabulates would put a secondary source where a primary one belongs.

The scoring framework

An ordinal 12-point scale, split evenly so that a drug with a strong basic rationale but no human data is not penalized for the gap:

The ITP distinction matters and is worth preserving in curation prose. The Interventions Testing Program uses genetically heterogeneous UM-HET3 mice, replicated across three sites, and is the reason a lifespan claim from it outranks a single-lab result.

The ranking

Note the difference between the two zeros. ACEi/ARB scores 0 on mortality because the studies exist and were predominantly negative; acarbose, rapamycin, methylene blue, senolytics, aspirin and NAC score 0 because the question was not assessed. Only the first is evidence against.

Drug × module matrix

Which hallmark modules each candidate has a documented mechanistic claim against, mapped onto kb/modules/. ✓ = curated as a treatment with target_mechanisms; · = claim documented in the source literature, not yet curated; blank = no applicable studies found.

The empty cells are informative and should not be filled in speculatively. The review records "No applicable studies" for many drug × hallmark pairs — SGLT2 inhibitors and methylene blue against epigenetics and stem-cell renewal, ACEi and NAC against most hallmarks — and the newer or less-studied agents are simply thinner than metformin and rapamycin, not proven inactive there.

Curation status

• [x] Survey aging-module treatment coverage (9 of 13 modules had none)
• [x] inflammaging — metformin, senolytics (D+Q)
• [x] gut_dysbiosis — metformin, acarbose
• [x] disabled_macroautophagy — aspirin, metformin, SGLT2 inhibition
• [x] mitochondrial_dysfunction — methylene blue, metformin
• [x] stem_cell_exhaustion — metformin preconditioning
• [ ] genomic_instability_aging — rapamycin (Werner fibroblasts), aspirin (UVB DNA damage)
• [ ] epigenetic_alterations — acarbose (PDX-1 methylation), aspirin (colonic hypermethylation)
• [ ] loss_of_proteostasis — rapamycin, metformin (UPR via AMPK/ERK1/2)
• [ ] cellular_senescence — extend beyond D+Q to metformin, rapamycin, methylene blue senomorphic effects
• [ ] telomere_attrition — metformin; thin evidence, may stay uncurated
• [ ] deregulated_nutrient_sensing — add acarbose (insulin/IGF-1) and SGLT2i (AMP/ATP ratio, mTORC1)
• [ ] Consider a Gerotherapeutics grouping over modules carrying these treatments

Curation notes

Effect direction is where the honesty lives. MODULATES is the right value more often than it is comfortable. Methylene blue improves mitochondrial respiration but its effect on reactive oxygen species runs in opposite directions depending on the substrate, so its link to Bioenergetic Decline and Oxidative Stress is MODULATES, and the entry carries a second evidence item quoting the unfavorable half of the result. Curating it as RESTORES with only the favorable quote would have been schema-valid and wrong.

directness: INDIRECT is the tool for the lifespan-to-mechanism leap. That canagliflozin extends median male lifespan by 14% does not show the benefit runs through autophagy. The lifespan result belongs in the entry — it is why the drug is worth curating at all — but as INDIRECT support, not as evidence for the mechanism link.

Watch evidence_source on human-cell studies. Much of the strongest metformin evidence comes from human T cells or adipose explants treated ex vivo. That is IN_VITRO, not HUMAN_CLINICAL, however human the donor. Xu et al. 2018 needs splitting into two items for this reason: the human-explant senolytic result is IN_VITRO, the mouse survival result MODEL_ORGANISM.

Sex dimorphism is a real finding, not a caveat to drop. Acarbose extends median male lifespan by 16–17% and female by 4–5%; canagliflozin extends male lifespan and not female at all. Where the source reports it, keep it in the evidence explanation.

One drug legitimately appears against several modules. Metformin is curated against inflammaging, macroautophagy, mitochondrial dysfunction, gut dysbiosis and stem-cell exhaustion. This is not duplication: Bharath et al. found the autophagic and mitochondrial actions run "largely in parallel" rather than one through the other, so they are separate mechanistic claims that happen to share a drug.

Open questions

• Should there be a gerotherapeutic module? The candidates share a target (aging) but not a mechanism, and a module is a conserved pathological process, not a drug class. A Grouping over the hallmark modules is the better fit; recorded as a checklist item above rather than assumed.
• ACEi/ARB is the awkward case. It scores 8, but its renoprotection is explained by hemodynamics rather than by gerotherapeutic action — the review says so explicitly. Curating it against an aging hallmark on the strength of its score would overstate the evidence.
• RTB101 / dactolisib was excluded from the ranking for not being FDA-approved, despite phase 2b/3 respiratory-infection results in older adults. If it is ever curated, the exclusion reason should be recorded so it is not mistaken for a negative result.

Sources

• Kulkarni AS, Aleksic S, Berger DM, Sierra F, Kuchel GA, Barzilai N. Geroscience-guided repurposing of FDA-approved drugs to target aging: A proposed process and prioritization. Aging Cell. 2022;21(4):e13596. PMID:35343051, doi:10.1111/acel.13596.
• Primary studies are cited per-claim in the module evidence blocks; see git log on kb/modules/ and the history records under history/modules/ for the per-session curation record.