Longevity Drugs: What Nature Aging Actually Found

When a headline says network intelligence has found «longevity drugs», it is easy to picture a list of familiar pills waiting for someone to try them. The Nature Aging study did not do that. It built a map for deciding which hypotheses deserve experiments.

The work is valuable precisely because aging does not fit inside one pathway. It connects genes, proteins, the hallmarks of aging, drug targets and gene-expression signatures. But no participant took a medicine, no heart attacks, frailty or survival were measured, and no new clinical indication emerged.

The translation for a reader is simple: use the paper to understand how research moves forward, not to choose what to buy. Between a computational prediction and a prescription sit years of validation, trials and risk assessment.

Editorial review: August 30, 2026. This article is educational and does not replace medical care. Do not start, stop or combine medicines because they appear in a geroscience publication.

The short answer

• The study began with 2,358 genes linked to longevity, age-related disease or aging pathways.
• It assigned 1,250 genes to 11 hallmarks: 860 to one and 390 to several, reflecting overlap between mechanisms.
• It evaluated 6,442 compounds that were approved or in clinical investigation by target proximity to those modules.
• It found 370 nearby one or more modules, meaning «may perturb them», not «are beneficial».
• Only 60 had CMap data for pAGE: 21 were positive, 23 negative and 16 inconsistent.
• The other 310 need expression data before pAGE can estimate direction.
• It is not a clinical trial: the results are falsifiable predictions.

Repurposing does not mean prescribing

Drug repurposing asks whether a known medicine might serve a new purpose. The advantage is practical: information already exists about targets, manufacturing, pharmacology and risk in the original use. That can shorten some early stages.

What does not automatically transfer is efficacy or the benefit-risk balance for the new purpose. A medicine acceptable for a serious disease may not have an acceptable risk as prevention in healthy adults. Dose, duration, population and combinations may also change.

The FDA's explanation of off-label use makes a useful distinction: a clinician may consider an unapproved use when medically appropriate, but the agency has not determined that the medicine is safe and effective for that purpose. «Known» does not mean «validated for everything».

How the map was built

Layer	What the team did	Main limitation
Genes	Started with 2,358 OpenGenes entries carrying different confidence levels.	Genetic association is not a validated therapeutic target.
Hallmarks	Connected 1,250 genes to 11 aging mechanisms.	Categories overlap and depend on current knowledge.
Interactome	Placed the genes in a protein-interaction network.	The network is incomplete and combines evidence from different contexts.
Drugs	Measured proximity between 6,442 compounds' targets and each module.	Closeness suggests perturbation, not direction or benefit.
Expression	Used CMap to ask how gene expression changes after a compound.	Cell line, dose and time may not represent a human organ.
pAGE	Estimated whether those shifts reinforce or counter age-associated signatures.	A favorable signature is not a clinical outcome.

The method is strong because it does not stop at a famous target. It integrates topology and direction. Its limitation is the same: every layer contains assumptions, incomplete data and models that still need testing outside the computer.

SHARP and pAGE without the jargon

SHARP is the name of the complete pipeline. First it asks whether a drug's targets sit near an aging module in the network. It then uses pAGE to ask which way the drug appears to move the module's gene expression.

Imagine an age-associated signature that raises some genes and lowers others. If a compound produces the opposite pattern, pAGE may consider it favorable. If it reinforces the pattern, pAGE may signal caution. This is a smart way to avoid saying «it touches the hallmark, therefore it is good».

pAGE does not measure how many years someone will live, is not a biological clock and cannot predict adverse effects by itself. CMap profiles depend on experimental cells, doses and timing. The paper itself discusses variation between cell lines and limited coverage.

What 370, 60, 21, 23 and 310 mean

Number	Correct meaning	Incorrect interpretation
370	Compounds with significant proximity to one or more modules.	370 drugs that slow aging.
60	Nearby candidates with enough CMap profiles to calculate pAGE.	60 medicines tested in humans.
21	Compounds with positive pAGE across analyzed confidence levels.	21 longevity treatments ready to use.
23	Compounds with negative pAGE, a potentially unfavorable computational signal.	23 medicines clinically proven to accelerate aging.
310	Proximity predictions without CMap data to resolve direction through pAGE.	310 hidden candidates that only need a dose.

The authors use «pro-longevity» for a positive pAGE direction. In context, that label describes a model result. Outside context, it can sound like clinical efficacy. A better translation is always «computational candidate with a signature that deserves validation».

Six gates stand between the computer and a recommendation

1. Reproduce the prediction: confirm it does not depend on one database, cell line or statistical choice.
2. Validate the mechanism: demonstrate the direction in relevant cells and tissues.
3. Measure dose and toxicity: separate a useful window from harmful exposure.
4. Choose a population: define the age, risk, disease or biomarker that justifies a trial.
5. Compare in humans: use randomization, placebo or standard care when appropriate.
6. Measure something meaningful: function, disease, events, quality of life or survival, not only expression.

This ladder explains why a computational paper can be excellent and still not change tomorrow's prescription.

A candidate list is not a shopping list

Familiar names attract headlines. Aspirin, metformin, rapamycin or a decongestant feel more accessible than an experimental molecule. Familiarity can lower our guard even though every medicine has its own indication and harms.

Aspirin is a useful reminder. The USPSTF says starting it for primary cardiovascular prevention between ages 40 and 59 with elevated risk should be an individual decision because net benefit is small, and recommends against starting for that purpose from age 60. Bleeding matters. Appearing in an aging network does not change that assessment.

Rapamycin follows the same logic: a powerful animal foundation does not create a human anti-aging indication. Our guide to rapamycin and longevity explains why an mTOR signal cannot become a do-it-yourself schedule.

Computational, preclinical, clinical and approved are not synonyms

Label	Question answered	Decision allowed
Computational	Which hypothesis deserves testing?	Prioritize research.
Preclinical	Is there an effect in cells or animals, and at what cost?	Design human studies if justified.
Early clinical	Is it tolerated and does it move the expected outcome?	Decide whether larger trials are warranted.
Confirmatory clinical	Does it improve relevant outcomes against a comparator?	Inform guidelines and indication.
Approved	Does benefit outweigh risk for a defined use?	Prescribe within the indication and label.
Off-label	Is there an individual clinical reason for an unapproved use?	Informed medical decision, not assumed general efficacy.

What a reader can do with this study

It can help you ask much better questions:

• Is this result computational, animal or human?
• Does the molecule have an indication for my problem or only plausibility?
• Which dose, tissue and duration were studied?
• Was the result a gene, a biomarker or something a person feels and values?
• Which risks and interactions sit outside the model?
• Is there an approved, better-studied alternative for the actual goal?

You can also start with your risk map. A review of biomarkers and function often reveals more concrete targets than «aging more slowly»: blood pressure, ApoB, glucose, strength, sleep or aerobic capacity. Proven interventions for those problems are not less advanced because they are less dramatic.

Conclusion: a good map is not the destination

The Nature Aging work improves candidate discovery by integrating a protein network with hallmarks and expression direction, producing predictions that can be disproved. That is good science.

Its output is not 21 anti-aging pills. It is 21 prioritized computational signals within the subset that had enough data, plus hundreds of open questions. The advance is knowing what to test next. The caution is refusing to act as if that testing has already happened.

FAQ

Does the study show that any drug extends human life?

No. It is a computational analysis that prioritizes hypotheses through protein networks and gene expression. It did not give the medicines to people or measure disease, function or survival.

What does drug repurposing mean?

It means studying a known medicine for a new use. Existing information about targets and safety may speed research, but the new use still needs its own clinical validation.

What is SHARP?

SHARP is the study pipeline. It combines proximity between drug targets and hallmark modules with a gene-expression metric called pAGE to rank research candidates.

What does pAGE measure?

pAGE estimates whether a compound-induced expression signature opposes or reinforces age-associated changes within a module. It is a model-dependent prediction, not a biological age or clinical outcome.

What does the finding of 370 compounds mean?

It means their targets were significantly close to one or more aging modules in the network. Proximity suggests potential perturbation, not benefit or safety.

Are the 21 positive-pAGE compounds longevity drugs?

No. They are candidates with a favorable transcriptional direction within the analysis. They need experimental validation and human trials before becoming prevention or treatment.

Should I take aspirin, metformin or rapamycin for longevity?

Not because they appear in a geroscience discussion. Each medicine has indications, contraindications and interactions. A decision should answer a concrete clinical problem, not a computational shortlist.

What is the difference between off-label use and a clinical trial?

Off-label use is an individual clinical decision to prescribe an approved medicine for an unapproved purpose. A clinical trial follows a research protocol, comparator, criteria and oversight to produce generalizable evidence.

Sources

• Gross B, Ehlert J, Gladyshev VN, Loscalzo J, Barabási AL. Network-driven discovery of repurposable drugs targeting hallmarks of aging. Nature Aging. 2026. Original article.
• Nature Aging. Mapping the network architecture of aging to identify repurposable drug candidates. 2026. Editorial analysis.
• López-Otín C, Blasco MA, Partridge L, Serrano M, Kroemer G. Hallmarks of aging: an expanding universe. Cell. 2023. Europe PMC 36599349.
• Subramanian A, Narayan R, Corsello SM, et al. Connectivity Map L1000 and gene-expression profiles. Cell. 2017. Europe PMC 29195078.
• Guarente L, Sinclair DA, Kroemer G. Human trials exploring anti-aging medicines. Cell Metabolism. 2024. Europe PMC 38181790.
• Espinoza SE, et al. Drugs targeting mechanisms of aging, NIA workshop proceedings. J Gerontol A. 2023. Europe PMC 37325957.
• FDA. Understanding unapproved use of approved drugs, off-label. Patient information.
• U.S. Preventive Services Task Force. Aspirin for primary prevention of cardiovascular disease. 2022. Recommendation.