AI Blood Tests That Predict Biological Age: How Close Are We?

Why Biological Age Matters More Than the Calendar

Two people who turn 60 on the same day can have radically different physiological conditions. One may be metabolically flexible, have the aerobic capacity of a 45-year-old, carry no chronic medications, and have biomarkers that all sit in the optimal range. The other may have hypertension, type 2 diabetes, early cognitive decline, and biomarkers clustering at the 75-year-old average. Their chronological age is identical; their biological age is not. The central goal of biological age research is to develop measurements that capture this divergence systematically, providing a more meaningful health metric than the calendar and, crucially, one that can track the impact of health interventions over time.

The clinical importance of this distinction becomes clear in two scenarios. In population health, biological age measurements allow identification of individuals whose rate of ageing is accelerated, flagging elevated disease risk years before conventional clinical thresholds are crossed. In personalised medicine, they provide an objective feedback metric for interventions including diet change, exercise programmes, sleep optimisation, and pharmaceutical or supplement protocols, allowing individuals and clinicians to test whether a given intervention is actually slowing or reversing measurable ageing processes. This is the "quantified self" idea applied to the fundamental process of biological ageing, and AI is the engine making it practically achievable.

The field has progressed through several generations of tools, each more informative than the last. Understanding the landscape requires distinguishing between blood chemistry models, epigenetic DNA methylation clocks, proteomic organ-clock approaches, and composite multiomic systems. As discussed in our article on biological age versus chronological age, each approach measures a different dimension of the ageing process and has different clinical strengths and limitations.

The Pioneer: Aging.ai and Standard Blood Chemistry

The earliest practical AI biological age tools emerged from the observation that routine blood chemistry panels contain substantial ageing information that conventional clinical interpretation largely ignores. In 2016, Alex Zhavoronkov and colleagues at the Insilico Medicine company (then based partly in collaboration with Russian research institutions and later prominent in AI drug discovery) published Aging.ai, a deep neural network trained on blood test results from over 60,000 individuals. The model used just 19 standard blood chemistry values to predict chronological age with a mean absolute error of approximately 6-7 years.

The 19 biomarkers included albumin (a marker of nutritional status and liver function), creatinine (kidney function), glucose, C-reactive protein (systemic inflammation), lymphocyte percentage, mean corpuscular volume (MCV), red cell distribution width (RDW), alkaline phosphatase, and several complete blood count parameters. These are all values measured in a standard complete metabolic panel and complete blood count, tests routinely ordered for almost any clinical presentation. This meant that Aging.ai could retrospectively analyse existing patient data without any additional testing, a powerful practical advantage.

The model's clinical validation showed that individuals whose predicted biological age exceeded their chronological age ("positive age acceleration") had higher rates of age-related disease and mortality in follow-up. More importantly for the precision medicine application, the model was sensitive enough to detect biological age changes over 6-12 month intervals in response to lifestyle interventions, something conventional clinical markers often miss. Zhavoronkov went on to use the platform extensively in self-experimentation, publishing data showing how his own biological age score responded to sleep deprivation, dietary changes, and supplementation protocols, establishing a template for n=1 longevity research that has become increasingly mainstream.

PhenoAge: Mortality Prediction from Nine Biomarkers

Morgan Levine at Yale (now at the Altos Labs research organisation) developed PhenoAge, published in 2018 in Aging Cell, using a more statistically rigorous approach. Rather than training the model to predict chronological age (which conflates biological ageing with time itself), Levine trained it on mortality risk, using data from the NHANES study population of over 11,000 Americans with long follow-up periods. The model was built to identify the combination of nine biomarkers that best predicted 10-year all-cause mortality, and the resulting score was then calibrated against chronological age to produce a biological age estimate.

The nine PhenoAge biomarkers are albumin, creatinine, glucose, CRP (log-transformed), lymphocyte percentage, MCV, RDW, alkaline phosphatase, and white blood cell count. The model has been extensively validated: individuals with a PhenoAge higher than their chronological age have significantly elevated risks of diabetes, cancer, cardiovascular disease, and all-cause mortality, with the associations holding even after adjustment for known confounders including smoking, BMI, physical activity, and socioeconomic status. PhenoAge acceleration is also associated with epigenetic ageing acceleration in the same individuals, suggesting it is measuring a real and biologically coherent ageing signal rather than statistical noise.

The clinical accessibility of PhenoAge is its major strength. It requires only values from a standard complete blood count and metabolic panel, costing roughly the same as a routine annual blood draw. Online calculators allow anyone with their blood test results to compute their PhenoAge in seconds. The limitation is sensitivity to acute perturbations: an acute infection, a recent intense exercise session, or a period of poor sleep can temporarily shift CRP and white cell count enough to inflate the PhenoAge score without reflecting true accelerated ageing. Interpretation therefore requires stable baseline conditions and ideally repeated measurements over time.

GlycanAge: Reading Immune Ageing in Sugar Chains

Gordan Lauc at the University of Zagreb took a different approach, focusing on the sugar coating of immunoglobulin G (IgG) antibodies. Glycans are complex sugar chains attached to proteins, and IgG glycosylation patterns change systematically with age in ways that reflect the inflammatory state of the immune system. Specifically, the ratio of pro-inflammatory to anti-inflammatory IgG glycoforms shifts with biological ageing, making the glycan profile a readout of what Lauc calls "inflammageing," the chronic low-grade inflammatory state that underlies most age-related disease.

GlycanAge, the commercial product developed by Genos based on Lauc's research, measures approximately 77 IgG glycan traits from a blood sample using liquid chromatography-mass spectrometry. The resulting biological age score has been validated in over 50,000 individuals across multiple European cohorts and shows strong associations with biological sex (women consistently score about 5 years younger than men in glycan age), lifestyle factors, and disease status. Importantly, GlycanAge shows meaningful responsiveness to interventions: a 2022 intervention study found that lifestyle changes including dietary improvement and exercise reduced GlycanAge by a mean of 6 years over 6 months in a cohort of overweight adults with elevated biological age scores.

The glycan approach has the advantage of measuring a specific, well-characterised biological mechanism (immune ageing and inflammageing) rather than a composite of markers with heterogeneous biological meanings. Its limitation is cost: the mass spectrometry analysis required is substantially more expensive than standard blood chemistry panels, and interpretation requires specialist knowledge. GlycanAge the company has made the test available to consumers at approximately 300-400 USD per test, positioning it as a premium longevity biomarker rather than a routine clinical tool.

Epigenetic Clocks and the DNA Methylation Revolution

The most biologically fundamental approach to biological age measurement uses DNA methylation, the chemical marking of cytosine bases in the genome that regulates gene expression. Steve Horvath at UCLA published the first pan-tissue epigenetic clock in 2013, showing that methylation levels at 353 CpG sites in the genome predicted age with a mean absolute error of 3.6 years across 51 different tissue and cell types. This was a breakthrough: a single molecular signature that tracked biological ageing uniformly across all tissues of the body.

Subsequent clocks have improved on Horvath's original in various ways. The Hannum clock (2013) was trained on blood specifically. The PhenoAge DNAmClock translates the blood chemistry PhenoAge concept into a methylation-based measurement. The GrimAge clock (2019) was trained directly on mortality risk rather than chronological age and includes seven methylation-based plasma protein surrogates alongside an estimate of smoking pack-years encoded in the methylation signature. GrimAge has shown the strongest associations with lifespan prediction of any current clock, with acceleration predicting time-to-death more accurately than any single clinical biomarker in multiple independent cohorts. DunedinPACE (2022) measures not biological age itself but the rate of ageing: how many months of biological time elapse per calendar year, with a score of 1.0 being average and scores above 1.0 indicating accelerated ageing.

Commercial access to epigenetic clocks has expanded significantly. Elysium Health offers its Index test, which uses a multi-clock approach including DunedinPACE. TruMe and myDNA offer methylation age testing from blood or saliva samples. The cost remains higher than blood chemistry approaches (typically 200-500 USD), and the information density is substantially greater. For those interested in tracking biological ageing longitudinally, methylation clocks currently represent the gold standard, though the field is moving rapidly toward organ-specific proteomic clocks as described in the next section.

Organ Age Clocks and Proteomic Profiling

A landmark 2023 paper in Nature by Tony Wyss-Coray's group at Stanford, using the SomaScan proteomics platform to measure nearly 5,000 proteins in blood plasma, introduced the concept of organ-specific biological ageing. By identifying protein clusters associated with specific organs (heart, brain, kidney, liver, immune system, muscle, fat, lung, and others), the researchers found that different organs age at different rates in the same individual, and that organ-specific age acceleration predicts organ-specific disease risk with striking specificity. A person whose brain ages faster than their other organs has elevated risk of Alzheimer's disease; accelerated heart age predicts cardiovascular events; accelerated kidney age predicts renal disease.

This organ clock concept is clinically transformative because it moves from a single whole-body biological age number to a profile of relative organ vulnerability, allowing targeted prevention strategies. It connects biological age testing to the AI drug discovery field, where proteomic signatures are being mined to identify drug targets for specific ageing processes in specific organs. Companies including Alkahest (a Grifols subsidiary), the Chan Zuckerberg Biohub, and several well-funded startups are developing clinical applications of proteomic organ clocks. Full proteomics panels remain research tools rather than consumer products at this stage, but targeted subsets are beginning to move toward clinical accessibility.

For consumers today, the most practical approach combines a blood chemistry-based biological age estimate (PhenoAge or equivalent, derivable from any standard annual blood panel) with periodic epigenetic clock testing (annually or semi-annually) to track trend over time. Adding GlycanAge provides specific immune ageing information. The combination of these three layers covers the major biological ageing axes measurable non-invasively and provides sufficient information to track the impact of lifestyle and clinical interventions on ageing rate, which is ultimately the most actionable output from the entire field.

Limitations, Regulatory Status, and What to Expect

Important caveats are worth emphasising before acting on biological age test results. First, these are probabilistic population-level tools that capture statistical patterns across thousands of individuals. A single measurement carries substantial uncertainty for any specific person. The signal-to-noise ratio improves with repeated measurements over time, making trend more reliable than any single data point. Second, none of these tests are currently approved by the FDA or similar regulatory bodies as diagnostic tools for any specific disease. They are wellness tools for now, sitting outside clinical medicine's regulatory infrastructure. Third, the optimal response to a high biological age score is unclear from the evidence: we know the score correlates with risk, but we do not have RCT-level evidence that reducing the score reduces clinical events in individuals (as opposed to populations).

The field is moving fast. Within a few years, comprehensive biological age profiling, combining blood chemistry, proteomics, epigenetics, and microbiome data into a unified AI-interpreted dashboard, is likely to be accessible at reasonable cost and begin entering clinical practice guidelines for preventive medicine. QuanMed AI is positioned at the intersection of these developments, providing the analytical infrastructure to make multiomic biological age data interpretable and actionable for individuals who want to engage seriously with the science of their own ageing process.