Network Medicine Framework Identifies Existing Drugs That Could Be Repurposed to Target Hallmarks of Aging
核心洞察
Northeastern and Harvard (搜索) researchers published a network medicine framework in Nature Aging that maps 1,250 aging-related genes onto the human interactome, revealing distinct "hallmark modules" for 9 of 11 hallmarks of aging.
The team introduced a novel metric called pAGE to determine whether drug-induced gene expression changes counteract or reinforce age-related transcriptional shifts, distinguishing pro-longevity from age-accelerating compounds.
The SHARP pipeline identified 370 drugs with significant proximity to aging hallmark modules; 21 showed positive pAGE as pro-longevity candidates, while 23 were flagged as potentially age-accelerating.
A team led by researchers at Northeastern University, in collaboration with Harvard (搜索), has developed a network medicine framework that systematically identifies existing drugs with the potential to be repurposed for targeting specific biological processes of aging. The findings, published Friday in Nature Aging, offer what the authors describe as "a roadmap toward actionable interventions that can be tested in cells, animals and eventually humans."
The study addresses a persistent bottleneck in longevity research: while thousands of genes have been implicated in age-related phenotypes, translating that knowledge into therapeutic interventions has proven difficult due to the multifactorial and interconnected nature of aging.
"We have a very big pile of genes related to aging," said Bnaya Gross, a postdoctoral researcher in Albert-László Barabási's lab at Northeastern and lead author on the study. "Networks allow us to organize them, saying, 'OK, it's not just a pile of genes. They are connected to each other. They form some sort of organization. It's not a random process.'"
Mapping the Hallmark Modules
The researchers began by querying the OpenGenes database, which manually curates gene-level annotations linking 2,358 genes to longevity, age-associated diseases, or pathways implicated in aging. From this resource, 1,250 genes were associated with at least one of the 11 established hallmarks of aging—biological processes ranging from genomic instability and telomere attrition to cellular senescence and mitochondrial dysfunction.
These 1,250 hallmark-associated genes were then mapped onto the human interactome, a comprehensive catalog of 524,156 experimentally validated binding interactions among 18,223 proteins. The analysis revealed that in 9 of the 11 hallmarks, the associated genes clustered into statistically significant connected subgraphs, or "hallmark modules" (z-score > 1.96). The remaining two hallmarks—loss of proteostasis (z-score = 1.74) and epigenetic alterations (z-score = 1.67)—also showed marginal significance, indicating nonrandom connectivity.
"If the genes were spread randomly, there is no way for a drug to specifically affect it because it's spread all over," Gross explained. "Our discovery is that aging genes are located in very specific areas, a very specific neighborhood, allowing you to find drugs that affect this neighborhood."
Further analysis using two complementary measures—separation and proximity—showed that these hallmark modules overlap and reside in the same network neighborhood, together forming a broader "longevity module."
Introducing the pAGE Metric
A key methodological advance of the study is the introduction of a transcription-based metric called pAGE (Pro-Age). While network proximity can establish a drug's ability to perturb a given hallmark module, it carries no information about whether that perturbation is beneficial or detrimental. The pAGE metric addresses this by quantifying whether drug-induced changes in gene expression reinforce or counteract documented age-related expression shifts.
The metric compares a drug's perturbation signature against a "longevity vector" encoding age-induced expression changes for 2,025 genes. Positive pAGE values indicate shifts toward younger expression patterns, while negative values suggest potential age-accelerating effects.
Together, network proximity and pAGE form the SHARP pipeline (Systematic Hallmark-based Aging Repurposing Pipeline), which first identifies compounds targeting hallmark-related subgraphs and then determines whether the induced transcriptional changes are pro-longevity or age-accelerating.
Validation Against Known Longevity Compounds
The researchers validated SHARP against three independent datasets. First, they examined 11 drugs that the Intervention Testing Program (ITP) found to prolong lifespan in mice. Of the 8 with available Connectivity Map (CMap) data, all 8 displayed positive pAGE and statistically or marginally significant proximity to at least one hallmark—yielding 100% sensitivity (95% confidence interval: 63%–100%).
Second, they tested 17 compounds currently undergoing clinical trials for healthy longevity, including metformin and sirolimus. Of the 9 with available CMap data, 8 displayed positive pAGE with significant or marginally significant proximity—88.9% sensitivity (95% confidence interval: 51.7%–99.7%).
Third, SHARP was evaluated against 10 compounds recently tested for lifespan and healthspan effects in mice in an independent parallel study. Among the 8 with CMap data, all compounds that significantly improved lifespan or healthspan showed significant or marginal proximity and positive pAGE—again 100% sensitivity (95% confidence interval: 47.8%–100%).
Drug Repurposing Candidates and Age-Accelerating Compounds
Applying SHARP across all hallmarks, the researchers identified 370 drugs with statistically significant proximity to one or more hallmark modules. Of these, 60 had CMap expression profiles enabling pAGE calculation: 21 displayed positive pAGE as pro-longevity candidates, 23 displayed negative pAGE as potential age-accelerating compounds, and 16 showed inconsistent pAGE values requiring further data.
Notable findings include aspirin, which was predicted to influence six hallmarks including altered intercellular communication and deregulated nutrient sensing. The over-the-counter nasal decongestant oxymetazoline (Afrin, Sinex) emerged as a candidate for improving cell-cell communication—a prediction now being validated through cell-line experiments with Harvard (搜索) colleagues.
Among the 370 candidates, 83 were classified as "network drugs"—compounds that perturb hallmark modules without directly targeting hallmark genes. These candidates "would be missed by approaches that do not consider the full network topology," the authors noted.
Limitations and Future Directions
The researchers acknowledged several limitations. Aging processes vary substantially across tissues and cell types, and the study primarily relied on MCF7 perturbation profiles due to broad CMap coverage. The pAGE metric does not yet account for dosage, nonlinear responses, or tissue-specific effects. Additionally, 310 of the 370 drug-repurposing predictions currently lack CMap data, though the authors estimate approximately 35% (about 108 drugs) of these may benefit longevity.
"This study does not provide a cure for aging, nor does it prove that any specific drug will extend human life," said Barabási, Distinguished University Professor of Physics at Northeastern who oversaw the study. "It offers a roadmap toward actionable interventions."
Gross noted that the methodology could eventually support personalized anti-aging strategies. "I think we are not there yet," he said, but "we are moving toward the era of preciseness."
The study underscores that effective longevity interventions will likely require combination therapies targeting multiple hallmarks simultaneously—a framework the network-based approach is well-positioned to support.