Deep Learning Blood Test Accurately Measures Cellular Senescence Burden and Predicts Age-Related Disease Risk

Researchers have developed an advanced deep learning-based biomarker known as the SASP Score, offering a novel method to evaluate the combined physiological effects of the senescence-associated secretory phenotype (SASP) circulating in human blood. This breakthrough addresses a long-standing challenge in geroscience: quantifying the systemic burden of senescent cells—often referred to as "zombie cells"—which accumulate with age, drive chronic inflammation, and accelerate the progression of multiple degenerative diseases.

Gathering the underlying data required to evaluate a person’s circulating SASP profile involves a routine blood draw, making the initial collection phase clinically accessible. However, interpreting this data has historically been exceptionally difficult. Bloodstream SASP is not a single entity; rather, it represents a complex mixture of numerous proteins secreted by diverse types of senescent cells located across various tissues and organs throughout the body. Furthermore, these proteins exhibit intricate, nonlinear relationships with one another and with various facets of human health. Consequently, measuring the entirety of the SASP as a cohesive, single-digit biomarker with genuine predictive utility has long eluded researchers.

To overcome these analytical hurdles, the research team turned to artificial intelligence, engineering a specialized deep learning algorithm they termed the SASP Score. According to the study’s authors, this algorithm successfully resolves two primary roadblocks in SASP analysis. First, it effectively accounts for the complex nonlinear interactions among different senescence-related proteins. Second, the model is robust enough to be deployed across multiple protein data-gathering platforms without requiring cumbersome data transformation steps.

The algorithmic foundation of the SASP Score relies on a Guided AutoEncoder with Transformer (GAET) architecture. This specific neural network design is exceptionally proficient at untangling complex, nonlinear relationships within large biological datasets and has previously demonstrated utility in the development of sophisticated aging clocks.

Methodological Rigor and the UK Biobank Dataset

The development of the SASP Score relied heavily on massive genomic and proteomic repositories, specifically drawing from the UK Biobank Pharma Proteomics Project (UKB-PPP). Out of a massive cohort of 54,219 participants available in the dataset, the researchers utilized high-quality data from 50,997 individuals. Prior to initial analysis, quality control measures led to the exclusion of only three proteins due to insufficient presence across the majority of samples.

Following an exhaustive review of existing scientific literature regarding cellular senescence, the research team curated a final panel of 38 distinct proteins to construct the biomarker. Many of these selected proteins—such as members of the CCL, CXCL, and interleukin (IL) families of inflammatory factors—are widely recognized in molecular biology for their roles in mediating senescence and tissue-level inflammation across a wide variety of cell types.

To train and validate the model properly, the researchers employed a standard machine learning split, randomly designating 85% of the dataset for model development and training, while strictly reserving the remaining 15% for independent validation to ensure the algorithm did not overfit the data.

Distinguishing the SASP Score from Traditional Biological Clocks

An important distinction emphasized by the study’s authors is that while chronological age was used during the model’s development phase to guide training, actual chronological age is deliberately excluded from the final SASP Score evaluations. Furthermore, the researchers are careful to clarify that because the metric focuses exclusively on tracking cellular senescence, it is not intended to serve as a comprehensive, general biological aging clock.

Nevertheless, when tested against UK Biobank data, individual SASP Scores demonstrated a strong correlation with chronological age—matching the correlative strength typically observed in established epigenetic and proteomic aging metrics such as PhenoAge, BioAge, and various other proteomic clocks designed to track healthspan and biological aging.

Beyond its baseline correlation with age, higher SASP Scores were consistently linked to numerous clinical markers of functional decline and physical deterioration. These indicators include a comprehensive frailty index, elevated blood pressure, measurable declines in pulmonary and cardiac fitness, slower walking speeds (gait speed), and reduced handgrip strength—all of which are classical clinical hallmarks of aging and declining physiological resilience.

Predictive Power for Age-Related Morbidity and Mortality

When controlling for conventional health-related confounding variables—such as chronological age, smoking status, alcohol consumption, blood pressure, and body mass index (BMI)—a higher SASP Score remained strongly predictive of a significantly elevated risk of all-cause mortality and age-related pathologies.

Statistical analysis revealed that individuals registering high SASP Scores faced roughly a 1.4 times greater likelihood of dying from any cause compared to those with low scores. When examining specific disease categories, the results varied: while certain types of cancer showed no statistically significant correlation or even an inverse relationship with high SASP Scores, other major chronic conditions exhibited profound associations. Specifically, the presence of an elevated total SASP Score was strongly correlated with a significantly higher prevalence and incidence of neurodegenerative and systemic diseases, including dementia, stroke, and chronic kidney disease. Notably, the SASP Score demonstrated superior predictive power for these specific conditions compared to evaluations of any single, isolated senescence-related protein.

The Role of Physical Activity: Insights from the MEDEX Study

To further validate the clinical utility and responsiveness of the SASP Score, the research team applied the metric to an entirely separate dataset derived from the MEDEX study, which was specifically designed to evaluate the physiological impacts of structured exercise.

Because the MEDEX study cohort was purposefully recruited to include healthy older adults free from specific, diagnosed age-related diseases, the baseline correlation between SASP Scores and chronological age was less pronounced in this group compared to the broader UK Biobank population. However, the longitudinal findings were striking.

Over the 18-month duration of the study, participants in the non-exercise control group experienced a statistically significant increase in their average SASP Scores. In stark contrast, the SASP Scores of individuals who engaged in regular, structured physical activity over the same 18-month period remained largely flat. This stabilization suggests that consistent exercise may exert a protective, therapeutic effect capable of braking or mitigating the systemic accumulation of senescence-associated secretory phenotypes over time.

Implications for Longevity Interventions and Future Research

The creators of the SASP Score emphasize that as a blood-based metric, the assessment provides a systemic and generalized overview of biological stress. The panel was intentionally designed to capture SASP proteins that are commonly expressed across highly heterogeneous populations of senescent cells throughout different tissues.

While a consistently elevated SASP Score serves as a statistically robust indicator of underlying, potentially severe long-term pathological conditions, it cannot pinpoint the exact anatomical location or specific identity of the disease driving the score. Consequently, the research team envisions this metric being utilized alongside traditional biological aging clocks as a valuable supplementary diagnostic tool. By capturing vital information regarding overall senescent cell burden, the SASP Score offers clinicians and researchers a rapid, cost-effective method to evaluate the real-time efficacy of emerging lifestyle modifications, therapeutic interventions, and senolytic pharmacological treatments aimed at clearing senescent cells and extending human healthspan.