AI Network Map Pinpoints Existing Drugs That Could Slow Aging Hallmarks
Harvard and Northeastern scientists used network medicine to screen 6,442 drugs against aging biology, revealing repurposing candidates for specific aging hallmarks.
Summary
Researchers from Harvard and Northeastern University built a computational framework that maps nearly 2,400 longevity-associated genes onto the human protein interaction network. By measuring how close 6,442 existing drugs sit to each of aging's core biological hallmarks, and by tracking whether those drugs shift gene expression in anti-aging directions, they identified promising drug-repurposing candidates. The approach also reveals the molecular mechanisms behind each candidate, making the findings testable rather than purely theoretical. This could dramatically accelerate the search for longevity medicines by mining drugs already proven safe in humans, rather than starting from scratch with new compounds.
Detailed Summary
Finding drugs that slow aging is one of medicine's hardest problems — not because we lack candidate genes, but because aging involves thousands of interacting genes across multiple biological systems simultaneously. No single drug target explains the whole picture, and that complexity has stymied drug development for decades.
To cut through this, a team led by Albert-László Barabási at Northeastern and colleagues at Harvard's Brigham and Women's Hospital built a network medicine framework. They placed 2,358 longevity-associated genes onto the human interactome — a map of all known protein-protein interactions — and showed that genes linked to each hallmark of aging cluster into distinct connected subgraphs, or 'hallmark modules.' This means each hallmark has a definable molecular neighborhood in the network.
With those neighborhoods mapped, the team calculated the network proximity of 6,442 known compounds to each hallmark module. Drugs sitting close in network space are more likely to modulate that hallmark's biology. They then developed a second metric called pAGE, which assesses whether a drug's gene-expression signature pushes hallmark-related genes in an anti-aging direction or worsens age-related expression patterns.
Combining proximity and pAGE produced a ranked list of repurposing candidates tied to specific hallmarks — and because the framework is mechanistically grounded, it explains why each drug might work. The authors describe the predictions as falsifiable, meaning they can be tested in the lab or clinic.
The implications are significant: rather than developing entirely new molecules, this approach could fast-track existing approved drugs into longevity trials. Caveats include the computational nature of the findings — no experimental validation is reported in the abstract — and the known incompleteness of the human interactome. Summary is based on the abstract only.
Key Findings
- 2,358 longevity genes were mapped onto the human interactome, forming distinct modules for each hallmark of aging.
- 6,442 existing drugs were scored by network proximity to aging hallmark modules, identifying repurposing candidates.
- A new metric, pAGE, measures whether a drug's gene expression shifts counteract known age-related changes.
- Combining proximity and pAGE yields mechanistically interpretable, testable drug-repurposing predictions.
- The framework could accelerate longevity drug development by leveraging compounds already proven safe in humans.
Methodology
This is a computational network medicine study using the human protein-protein interactome as its scaffold. Longevity-associated genes were mapped to identify hallmark-specific subgraphs, and 6,442 compounds were scored using both network proximity and a novel transcriptomic metric (pAGE). No in vitro or in vivo experimental validation is described in the abstract.
Study Limitations
The study is purely computational; experimental or clinical validation of the top candidates is not reported. The human interactome remains incomplete, which could affect proximity calculations and introduce bias. This summary is based on the abstract only and may omit methodological details, validation experiments, or nuances present in the full paper.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: