Geroscience in 2025: The Expert Roundup

2025 was a good year for geroscience, marked by rapid strides and critical milestones. Yet, the path wasn’t always smooth: progress in some areas lagged, research hit dead ends, and familiar bottlenecks persisted.

Against this backdrop, we asked five prominent geroscientists to share their perspectives on the highs and lows of 2025, along with their outlook for 2026. We are proud to have covered many of the breakthrough papers they highlight, and we remain committed to keeping you at the forefront of science’s most important field in the year ahead.

Where did geroscience exceed your expectations in 2025, and where did it underperform or stall? Has this changed your long-term outlook for the field’s progress?

Steve Horvath, Principal Investigator at Altos Labs

Geroscience exceeded my expectations through the rapid integration of AI and foundation-model thinking into aging biology. Large language-style models are now being adapted to molecular data, enabling representations that generalize across cohorts, species, tissues, and platforms. A striking example is CpGPT: A Foundation Model for DNA Methylation by Lucas Paulo de Lima Camillo.

CpGPT illustrates how such models can already generate improved versions of established epigenetic clocks, e.g. new versions of GrimAge that appear to outperform existing epigenetic clocks. This represents a qualitative shift: from handcrafted biomarkers to self-improving aging models. In this context, I also want to mention the paper on BioLearn by Albert Ying and Vadim Gladyshev in Nature Aging: “A unified framework for systematic curation and evaluation of aging biomarkers.”

Where the geoscience field disappointed me was in human rejuvenation trials. In particular, plasmapheresis, despite compelling animal data and strong narratives. A couple of published studies largely failed to produce convincing improvements in epigenetic clocks or other molecular aging markers in humans. For example, the negative or null findings reported by Borsky et al. underscore how difficult it remains to translate systemic interventions into durable molecular rejuvenation.

George Church, Professor at Harvard Medical School & MIT

Previously neglected categories of aging-reversal targets exceeded expectations as they matured from 2024. For example, the ATPIF1 gene product (mitochondrial ATPase inhibitory protein) has renewed interest in reactive oxygen species (ROS) and led to the use of a repurposed ischemia drug. We also saw progress in engineered enzymes designed to break extracellular crosslinks, such as the glycoxidation end-product N-epsilon-(Carboxymethyl)lysine (CML).

Clinical trials for dietary supplements underperformed and were underfunded. Similarly, overfunded multi-billion-dollar startups seemed to lag behind, lapped by nimble, creative NewCos. The long-term outlook is bright as more research teams commit seriously to the multiple targets and tissues of aging (polypharmacy).

Andrea Maier, Professor in Medicine at the National University of Singapore

In 2025, geroscience exceeded expectations in the precision and scalability of biological age measurement. We finally moved from “promising biomarkers” to clinically deployable, interoperable aging phenotypes combining multi-omic clocks, digital mobility signatures, and organ-specific functional reserves. The integration of cardiovascular, immune, and musculoskeletal risk trajectories into actionable platforms is a milestone I did not think we would reach this soon; this enables precision geromedicine.

Where the field clearly underperformed was in late-stage interventional trials. The pipeline is rich, but translation slowed due to inconsistent phenotyping and regulatory uncertainty around composite aging endpoints. Several highly anticipated therapies showed safety but only modest clinical effect sizes, an expected but still sobering reminder that human aging biology is multi-dimensional.

These dynamics have not dampened my long-term optimism. If anything, 2025 reinforced that precision multimodal interventions, not single gerotherapeutics, are the path forward. We are entering the era in which geroscience becomes truly medical: measurable, monitorable, and modifiable.

Matt Kaeberlein, CEO of Optispan

One thing that impressed me in 2025 was how much traction geroscience gained outside the usual scientific circles. There’s been a clear shift in awareness among industry leaders and policymakers.

At the A4Li Summit in April, for example, Dr. Mehmet Oz, now Administrator of the Centers for Medicare & Medicaid Services, explicitly referenced geroscience and the Hallmarks of Aging in his remarks. That would have been unthinkable not long ago.

We now have a Longevity Science Caucus in the U.S. House, and similar initiatives are emerging internationally. Geroscience is increasingly part of a broader conversation about healthcare sustainability, prevention, and long-term outcomes, and that momentum is encouraging.

On the flip side, 2025 was another year where no intervention convincingly outperformed rapamycin, let alone caloric restriction, in terms of effect size on aging biology. While there has been meaningful progress translating geroscience principles into clinical practice, particularly around lifestyle-based interventions, and while a few promising candidates are moving through regulatory pipelines, I don’t see strong evidence that we’re close to large-effect interventions that can substantially slow or partially reverse aging. There’s no shortage of hype, but the data simply haven’t caught up yet. That hasn’t dampened my long-term optimism, but it has reinforced the need for rigor and innovative discovery science.

Oliver Medvedik, Chief Science Officer at Lifespan Research Institute

There wasn’t really one critical moment in 2025 in this field for me; rather, it has been the culmination of progress that I have witnessed over the years. There has been an explosion of interest in this field from students, along with biotech companies that are specifically focusing on aging that just wasn’t there 20 years ago.

This has been driven by both new discoveries, such as the identification of the role of epigenetic alterations in aging, along with new technologies such as CRISPR and cellular reprogramming factors. However, despite these advances, it is sobering to realize that we don’t have any therapeutic interventions available that significantly outperform calorie restriction when it comes to extending lifespan.

Which paper, experiment, or trial from 2025 do you consider the most influential, and why? You’re welcome to name a runner-up.

Steve Horvath

Based on both scientific impact and personal correspondence, the most influential paper of 2025 appears to be: Lei, J., et al. Senescence-resistant human mesenchymal progenitor cells counter aging in primates. Cell, 2025. This study crossed a critical boundary by demonstrating that engineered, senescence-resistant human cells can improve aging-related phenotypes in non-human primates.

The paper that most altered my personal thinking was Jayne, L. et al. A torpor-like state in mice slows blood epigenetic aging and prolongs healthspan. Nature Aging, 2025. By showing that reduced body temperature and torpor-like physiology slow epigenetic aging, this work clarified that epigenetic clocks are deeply responsive to thermometabolic states. It taught me something fundamental about what the methylome is sensing. Overall, 2025 was a landmark year for understanding the mechanisms underlying epigenetic aging, with multiple breakthrough studies substantially advancing the field.

George Church

The grand challenge of highly specific therapeutic delivery – required for lower doses and lower toxicity – saw numerous advances. For example, intravenous injection of a new AI-designed AAV vector showed 280-fold higher brain transduction and 50-fold lower liver transduction compared to the previous best in this category (AAV9).

This underlines how AI applications for protein engineering and target discovery are arguably the most impressive and rapidly improving areas of the field. Their influence is beginning to impact powerful de-aging and anti-cognitive decline strategies. The use of multiplex libraries with molecular barcodes enables an immediate path from millions of AI designs into primate testing for the cost of a single animal, skipping often-misleading in vitro and rodent phases or under-powered ‘single-design-at-a-time’ approaches.

Andrea Maier

The single most influential contribution in 2025 was ‘From geroscience to precision geromedicine’ led by Guido Kroemer. This was not just a review or a position piece; it represented a collective, cross-disciplinary manifesto from leaders across molecular biology to translational medicine. The unusually broad authorship functions as a de facto consensus about how we should now conceptualize aging: as a systemic, network-mediated process driven by “gerogenes” (aging-promoting pathways) and modulated by “gerosuppressors,” interacting dynamically with environment and behavior.

We describe a shared framework that has the power to unify previously fragmented subfields – proteostasis, immunosenescence, metabolic dysregulation, and organ cross-talk – under a clinically relevant, mechanistic paradigm. It does so not by rehashing old “hallmarks,” but by advocating for a precision geromedicine approach: identifying actionable pathways, matching interventions to individual molecular “aging signatures,” and integrating multi-omic, functional, and clinical data to inform real-world treatments. It means geroscience is no longer just a promising theory but feeds into the translational, precision-medicine discipline.

A close runner-up was Tony Wyss-Coray’s organ-specific proteomic clock study, which offered the most granular evidence to date that biological age is fundamentally organ-resolved. By quantifying distinct proteomic trajectories for brain, liver, kidney, muscle, and vasculature, the study showed that individuals do not age uniformly and that interventions will need organ-specific endpoints and timing. Clinical data were already available; now we also have biological evidence. This has immediate implications for trial design, risk stratification, and clinical implementation.

Matt Kaeberlein

It’s hard to single out one “most influential” result in a field as broad as geroscience. From a pragmatic, real-world perspective, I was particularly impressed by the LinAge2 paper from Jan Gruber’s group. It represents one of the first aging clocks that is both easy to implement and grounded in actionable, clinically validated measures, with a clear link to future mortality risk. That combination of being practical, interpretable, validated, and clinically relevant is exactly what the field needs right now.

As a runner-up, I’d point to the FDA approval of rapamycin for veterinary use under the brand name Felycin-CA1. Although the indication – heart disease in cats – is not a pure geroscience endpoint, the scientific rationale for using rapamycin in age-related cardiac decline comes directly from geroscience research in mice and dogs. This approval represents a critical regulatory milestone and may provide a template for how existing gerotherapeutics can be responsibly repurposed, particularly in companion animals.

Oliver Medvedik

Not a paper or trial necessarily, but rather the progress of organizations such as the Biomarkers of Aging Consortium in promoting the research and adoption of a variety of biomarkers of aging in determining biological age. As a runner-up, I know this is biased, but I’ll add the recent paper from the Sharma Lab, “Cell-Surface LAMP1 is a Senescence Marker in Aging and Idiopathic Pulmonary Fibrosis.” I thought it to be a well-crafted paper that identifies a potentially very useful target for senescent cells.

Thinking back to what you believed about aging biology on January 1, 2025: did any results this year genuinely surprise you or make you update your priors?

Steve Horvath

I was genuinely surprised by evidence showing that acute, high-intensity exercise can transiently reduce epigenetic age measured in saliva within 90 minutes. In professional soccer players, 90 minutes of intense match play produced measurable reductions in epigenetic age estimates, which rebounded within a day.

This finding was surprising not because exercise is beneficial but because it revealed how rapidly and reversibly the DNA methylome can respond to extreme physiological stress. It forced me to update my assumptions about temporal stability and reinforced the importance of sampling timing, tissue choice, and acute exposures when interpreting epigenetic clock data. This has consequences for geroscience clinical trial design and biomarker deployment. Arkadi Mazin wrote a very nice article about this study.

George Church

Since aging impacts most or all body tissues, the tendency to target the liver would seem to leave many holes. However, in addition to the intravenous targeting mentioned above, another ray of hope is that seemingly non-secreted (ergo cell-autonomous) therapies – like transcription factors (OSK) and RNA splicing factors (SRSF1) – can nevertheless (surprisingly) impact total body function. This is reflected in preclinical (animal) survival curves even if administered very late in life.

Such survival curves are not required (or even welcome) for FDA approval, but they are good signs that the field is getting closer to general and core aging mechanisms, not just biomarkers or one-off symptoms.

Andrea Maier

The study by Li et al “Multiomics and cellular senescence profiling of aging human skeletal muscle uncovers Maraviroc as a senotherapeutic approach for sarcopenia” builds the first senescence atlas of human skeletal muscle by applying single-nucleus multi-omics (RNA + chromatin accessibility) to over 50,000 muscle-derived nuclei from young and older donors. Aging muscle harbors widespread and highly heterogeneous senescent cells across multiple cell types (muscle stem cells, fibro-adipogenic progenitors, endothelial and smooth muscle cells), with considerable variation in transcriptomic and epigenomic senescence signatures.

They also map the senescence-associated secretory phenotype (SASP) of these cells, revealing both shared and cell type-specific SASP factors. Importantly, the study identifies drugable SASP components and demonstrates in mice that the HIV drug Maraviroc can mitigate age-associated muscle mass decline (sarcopenia), suggesting a senomorphic (SASP-modulating) therapeutic strategy. That moves senescence from a theoretical hallmark into an actionable therapeutic target in humans. Trials are needed!

Matt Kaeberlein

My thinking about aging biology has shifted based on several talks and conversations from the Global Conference on Gerophysics held at NUS this year. There seems to be a growing consensus that species’ maximum lifespan may be harder to overcome in humans than we’ve previously anticipated.

The work of Peter Fedichev and others in this area is compelling and suggests that the mechanisms underlying the maximum-lifespan barrier in humans may be fundamentally different than the mechanisms the field has been focused on for several decades, which may be more able to impact healthspan and population median lifespan in humans. It’s a reframing that doesn’t diminish the importance of geroscience, but it does sharpen our expectations and suggests, once again, the importance of a return to discovery science for this field.

Oliver Medvedik

Nothing has really surprised me, nor have any results from this year have had me update any major priors, as of yet. I still view aging as a consequence of the inability of various intrinsic cellular repair systems to keep up with the pace of damage accumulation/cellular entropy of varying sorts until homeostasis has collapsed. Species and populations of cells have evolved mechanisms that effectively solve this problem, whereas individual cells and somatic cells in tissues, for example, have not. Therein lies the rub.

What do you expect from 2026 in terms of scientific breakthroughs, clinical progress, and the regulatory or business climate for geroscience?

Steve Horvath

I expect the unexpected. We have entered a phase of accelerating convergence, where biological insight, data scale, and computational power are reinforcing one another. Datasets are becoming unprecedented in both size and depth, enabling questions that were simply inaccessible a few years ago. At the same time, AI models are becoming reusable and transferable.

Clinically, a new generation of well-powered, biomarker-informed trials is coming into view. Scientifically, this convergence is reshaping how we think about aging as a modifiable process. It has truly never been a more exciting time to work in geroscience.

George Church

I expect 2026 to see more cases of dramatic shortening in the FDA approval process, representing a general sea change. Examples include the first iPSC-derived therapy to receive FDA clearance for a Phase 3 trial (the Fertilo product from Gameto), and the sprint from disease diagnosis to FDA-approved cure in just 7 months for ‘Baby KJ’ (using kayjayguran abengcemeran, an intravenous LNP-delivered CRISPR base editor) [4].

Systems medicine and AI, along with the multiplex primate testing mentioned above, should help catalyze this trend via better initial designs for safety and efficacy. We expect better antibodies and other categories of binders to create improved agonists, antagonists, targeting mechanisms, and delivery systems.

2026 could also be a turning point for life-extending cell and organ therapies, as 2025 marked by far the longest xenotransplant (pig-to-human) survival, keeping a patient free of kidney dialysis for 271 days. The ARPA-H FRONT program (Functional Repair of Neocortical Tissue) recognizes this and seeks to catalyze progress in the most challenging and inspiring area of aging-relevant repair: specifically, the brain regions impacted by stroke and other late-onset maladies.

Andrea Maier

In 2026, I expect the field to pivot decisively from generalized gerotherapeutics to stratified, mechanism-matched, and gerodiagnostic-matched therapeutics. We will see the first trials powered on organ-specific biological age reversal – specifically muscle, immune, and vascular – and regulators will increasingly accept multi-dimensional aging outcomes when linked to validated risk reduction.

Scientifically, I anticipate breakthroughs in the safety of human in vivo cellular reprogramming, particularly regarding transient, tissue-restricted epigenetic reset strategies.

Clinically, we will see the expansion of healthy longevity medicine (academically termed precision geromedicine) into mainstream healthcare systems, as insurers recognize the value of quantifying and modifying risk for age-related diseases and functional decline decades earlier. The business environment will reward companies that integrate gerodiagnostics, gerointerventions, and longitudinal monitoring, rather than those chasing single-product silver bullets based on no evidence.

Matt Kaeberlein

I think there’s a very high likelihood that we’ll see the first FDA approval for a pure geroscience indication in companion animals. Loyal, for example, appears to be on track for conditional approval for lifespan extension in dogs, and that would be a watershed moment for the field. Success there would lower regulatory barriers and create a clearer pathway for others to follow.

Clinically, I expect continued acceleration in the integration of AI into healthspan medicine. We’ll certainly see advances in imaging analysis and predictive biomarker algorithms. What I’m most excited about, though, is the emergence of personalized healthspan agents: AI systems capable of integrating medical records, diagnostics, and wearable data to provide continuous, individualized feedback. If done well, these tools have the potential to make geroscience actionable at scale while keeping the focus on evidence rather than hype.

Oliver Medvedik

I find it impossible to have expectations when it comes to scientific breakthroughs or clinical progress. That said, I do have some rather modest expectations for the regulatory landscape in 2026 that should be attainable and, if enacted, quite possibly would lead to the largest boost in the field of geroscience to date.

Simply put, if we can identify biomarkers of aging that reliably predict biological age, and thus reliably couple that with increased risk of morbidity, then we can identify a sub-population that can be characterized as having “accelerated aging.” This would enable us to then recognize this particular group as having a true disease, as medically defined, rather than considering all of aging as a disease state, which the medical establishment is averse to do.

Having “accelerated aging” thus defined would, in my opinion, open the floodgates towards more streamlined clinical trials that encourage the development and adoption of novel classes of drugs that target the root causes of aging. Initially prescribed for people who fit the category of “accelerated aging,” i.e. having a biological age that is maybe 5-10 years older than their predicted chronological age, these classes of drugs would in turn be predicted to have efficacy in people who exhibit “normal aging” as well. This, to me, is the best way to get through the present bottleneck of translating the findings of geroscience into the clinic.