Most chronic diseases don’t begin with obvious symptoms or dramatic warning signs. Instead, they develop quietly over many years, as small changes accumulate in the body. A new perspective from researchers at the Buck Institute for Research on Aging notes that modern medicine often waits until disease is well underway and argues that new technologies could help detect risk much earlier, when prevention may be most effective.

The perspective, aptly titled “We Wait for Disease to Shout. What if We Listened When Biology Whispered?” introduces the concept of the “long tail” of biology. Rather than being caused by a single factor, most diseases and aging-related conditions develop from the combined impact of many small influences, including genetics, lifestyle, environmental exposures, sleep patterns, stress, and changes in the gut microbiome. Over time, these subtle shifts can gradually weaken the body’s resilience and increase the risk of chronic disease.

“By the time many diseases are diagnosed, the body has often been drifting off course for years,” said Nathan Price, PhD, Buck Institute professor, co-director of the Buck’s Center for Human Healthspan and senior author of the paper. “We now have the opportunity to detect those early changes by tracking what’s normal for each individual and noticing when biology starts to move in the wrong direction.”

The researchers highlight how diseases such as type 2 diabetes, heart disease, and neurodegenerative disorders often begin developing long before symptoms appear. For example, in type 2 diabetes, biological changes related to inflammation, metabolism, and insulin function can occur 10 to 15 years before blood sugar levels rise enough to trigger a diagnosis. The authors argue that catching these early warning signals could open the door to interventions that help delay or even prevent disease.

To make this possible, the perspective proposes a new personalized framework that treats each individual as their own biological reference point. By tracking changes over time, rather than comparing someone to population averages, researchers believe it may be possible to identify subtle shifts that signal increased risk.

Advances in health technology are making this approach increasingly realistic. Wearable devices can now continuously track heart rate, sleep, activity, and other physiological signals, while modern laboratory techniques allow scientists to measure thousands of biological markers from simple samples such as blood, saliva, urine, or even breath. Combined with artificial intelligence tools that can analyze complex patterns, these technologies could help translate large amounts of data into meaningful, personalized insights.

“Medicine has traditionally focused on treating disease after symptoms appear,” said Noa Rappaport PhD, lead author of the paper and an associate research professor at the Buck Institute. “Our goal is to shift toward protecting health by identifying risk earlier and understanding how each person’s biology changes over time.”

The authors also emphasize that major challenges remain. “Advanced biological testing can still be expensive, and healthcare systems are largely designed to treat illness rather than monitor long-term health,” said Lee Hood, MD, PhD, distinguished professor and co-director of the Buck’s Center for Healthspan. “Ensuring broad access to preventive technologies will be critical to preventing new health disparities. In addition, regulatory systems will need to adapt to evaluate new approaches that rely on personalized data and AI-driven analysis.”

Despite these challenges, the researchers say the tools needed to transform prevention are rapidly emerging. By combining wearable sensors, advanced biological measurements, and artificial intelligence, they envision a future in which healthcare focuses not just on treating disease, but on preserving health throughout life.

Citation: We Wait for Disease to Shout. What if We Listened When Biology Whispered?

DOI: 10.1016/j.cels.2025.101509  

Additional Buck Institute coauthor: Annalise Schweickart also contributed to the work.

COI: Nathan Price is chief scientific officer at Thorne and has a profit interest in the company. He also serves as an advisor to the Institute for Healthier Living, Abu Dhabi, and various companies where he has equity, including Vitaliti, Rue Four, ProPetDx, and Sera Prognostics.

Acknowledgements: This work was funded by an award from the Proactive Health Office of the Advanced Research Projects Agency for Health (ARPA-H) to the Personalized Analytics for Transforming Health (PATH) Project, the NIH NIA T32 AG000266 grant for Training in Basic Research on Aging and Age-Related Disease, and National Institutes of Health (NIH) grant no. U19AG023122 528

About the Buck Institute for Research on Aging

At the Buck, we aim to end the threat of age-related diseases for this and future generations. We bring together the most capable and passionate scientists from a broad range of disciplines to study mechanisms of aging and to identify therapeutics that slow down aging. Our goal is to increase human health span, or the healthy years of life. Located just north of San Francisco, we are globally recognized as the pioneer and leader in efforts to target aging, the number one risk factor for serious diseases including Alzheimer’s, Parkinson’s, cancer, macular degeneration, heart disease, and diabetes. The Buck wants to help people live better longer. Our success will ultimately change healthcare. Learn more at: 

Journal

Cell Systems

Funder

ARPA-H,
NIH/National Institute on Aging,
NIH/National Institutes of Health

DOI

10.1016/j.cels.2025.101509

Frailty—a multidimensional syndrome characterized by diminished strength, endurance, and physiological function—is notoriously challenging to predict accurately. Its presence significantly elevates the risk of adverse health events such as falls, hospitalization, and mortality, particularly among elderly individuals with CKD. Traditional predictive models often rely on cross-sectional data and superficial correlations, which while informative, fail to fully capture the nuanced causal relationships that drive frailty progression. The study under discussion addresses this critical limitation by harnessing causal feature learning, a method that goes beyond association to identify features with direct influence on patient outcomes.

What sets this research apart is its commitment to individualized prediction. Recognizing that frailty manifests differently across patients due to genetic, environmental, and comorbid condition variabilities, the AI framework is designed to personalize risk profiles. Embedded causal feature extraction allows the model to discern which factors hold genuine predictive power for a given individual, such as specific biomarkers, clinical history elements, or lifestyle parameters. This granularity is essential for developing interventions that are not only effective but also patient-centric and ethically sound.

The methodology integrates advanced machine learning architectures that perform knowledge distillation—a process where a complex, highly accurate model (the “teacher”) transfers its learned knowledge to a simpler, more interpretable model (the “student”). This approach ensures that the final predictive framework is both powerful and usable in real-world clinical environments. Clinicians can thus benefit from transparent decision-support tools without sacrificing predictive precision, bridging the notorious “black box” gap that often hampers AI’s clinical adoption.

Furthermore, the causal learning backbone enhances the model’s robustness against confounding variables and biases commonly encountered in medical datasets. By identifying true causal relationships rather than merely correlational patterns, the AI-driven framework promises resilience when applied to diverse patient populations and external validation cohorts. This addresses a critical bottleneck in medical AI—generalizability—which is paramount for any tool aiming for widespread clinical implementation.

The frailty prediction initiative detailed in this protocol also features a dynamic intervention component. Leveraging the rich causal insights, the system not only forecasts frailty risk but actively informs tailored therapeutic strategies. These interventions might include optimized pharmacological regimens, personalized nutrition plans, or specific physical rehabilitation protocols that align directly with each patient’s unique frailty determinants. This adaptive feedback loop exemplifies the shift toward precision medicine, wherein AI systems do not merely assess risk but empower proactive, individualized care planning.

Mounting evidence underscores the heavy toll of chronic kidney disease on elderly populations, where frailty accelerates morbidity and complicates management. By embedding AI at the intersection of nephrology and geriatric care, this research ventures into uncharted territory. It aims to capture the multifactorial etiology of frailty with unprecedented clarity, enabling healthcare providers to anticipate and mitigate decline before clinical deterioration occurs. This proactive stance could substantially reduce healthcare costs while improving quality of life for some of the most vulnerable patients.

Clinical datasets feeding the AI framework are meticulously curated, integrating longitudinal data from electronic health records, laboratory results, imaging, and patient-reported outcomes. The large-scale, multi-center nature of these datasets enriches the AI’s learning capacity and supports the extraction of reliable causal signals amidst noise and variability. This extensive data fusion epitomizes modern health informatics, where synergy between diverse data types fuels next-generation predictive analytics.

Importantly, the research team has planned rigorous validation phases, encompassing retrospective analyses and prospective clinical trials. Such stringent testing is vital to ensure the system’s efficacy and safety before deployment. Ethical considerations also accompany this innovation, with explicit attention to patient consent, data privacy, and algorithmic transparency. These safeguards promote trust among both patients and practitioners, a key factor for successful AI integration in sensitive areas like frailty assessment.

The potential impact of this AI-powered prediction and intervention framework extends beyond nephrology and geriatrics. By demonstrating how causal inference and knowledge distillation can coalesce in personalized medicine, the study sets a precedent for analogous applications in other chronic conditions where frailty and functional decline are prevalent, such as chronic obstructive pulmonary disease, heart failure, and neurodegenerative diseases.

As AI continues to reshape healthcare landscapes, this protocol highlights the critical symbiosis between cutting-edge data science and clinical insight. The collaborative effort between computer scientists, nephrologists, geriatricians, and bioinformaticians has produced a model that respects the complexity of human biology while offering scalable solutions to pressing clinical challenges. Such multidisciplinary synergy is a hallmark of future-proof innovations destined to thrive in the 21st-century healthcare ecosystem.

In summary, the advent of an AI-driven individualized frailty prediction and intervention framework represents a transformative advancement for elderly patients grappling with chronic kidney disease. Through causal feature learning and knowledge-distillation, the framework achieves a nuanced understanding of frailty drivers, empowering personalized preventative strategies and precision care. Beyond its immediate clinical promise, this research exemplifies how sophisticated AI methodologies can be responsibly harnessed to tackle multifaceted medical problems, fostering healthier aging populations worldwide.

Subject of Research: Development of an AI-driven individualized frailty prediction and intervention framework for elderly patients with chronic kidney disease using causal feature learning and knowledge-distillation-based modeling.

Article Title: Protocol for development of an AI-driven individualized frailty prediction and intervention framework for elderly patients with chronic kidney disease: causal feature learning and knowledge-distillation-based modeling study.

Article References:
Chang, J., Hu, J., Cao, Y. et al. Protocol for development of an AI-driven individualized frailty prediction and intervention framework for elderly patients with chronic kidney disease: causal feature learning and knowledge-distillation-based modeling study. BMC Geriatr (2026). https://doi.org/10.1186/s12877-026-07143-0

Image Credits: AI Generated