Brain Age Gaps Map Aging Biology, Not Anti-Aging Cures

A brain scan that looks older than its owner may reveal something real about aging biology. It may also reflect a model's training data, the MRI measures it used, and the statistical quirks of predicting age. The useful conclusion is narrower than the marketing: brain age gaps can help researchers map pathways and prioritize drug targets. They do not show that lowering a brain-age score will prevent dementia, extend life, or reverse aging.

That distinction matters as age clocks move from academic papers toward wellness dashboards. A biological age gap usually means predicted age minus chronological age. A positive brain age gap means an MRI model judged the brain older than expected for the person’s calendar age. It is a constructed biomarker, not a diagnosis and not a direct measurement of the rate at which a person is aging.

Brain age has genetic signal, but it is not one master aging score

Fan Yi et al.'s 2025 Science Advances study built a brain age gap from T1-weighted MRI in UK Biobank. The researchers trained deep-learning models in 29,097 healthy participants, selected a 3D vision-transformer model after validation in 3,227 healthy people and 6,637 people with brain disorders, and also checked performance in the ADNI, PPMI, and IXI datasets. They then conducted genetic analyses in 31,520 healthy genotyped participants.

The study found that brain age gap had a meaningful polygenic component, with SNP heritability of about 0.2. That does not mean 20% of anyone’s brain aging is fixed by genes. It means common genetic variants explained roughly one-fifth of variation in this particular MRI-derived trait within the study population. A separate 2025 genome-wide brain-age analysis by Jawinski et al. identified 59 associated loci and linked higher brain age to mental and physical health phenotypes. Together, the papers argue that brain age captures more than scanner noise. They do not establish a single biological process called "brain aging."

The larger systems picture is similarly mixed. In a 2024 Nature Aging atlas by Wen et al., researchers estimated biological age gaps across nine organ systems in 377,028 European-ancestry UK Biobank participants. They used non-overlapping samples for genome-wide association analyses (154,774 participants) and Mendelian randomization analyses (222,254 participants). Across the atlas, they reported 393 locus–age-gap pairs; 11 loci met genome-wide significance for brain age gap. Annotation linked brain-age loci to brain, metabolic, neurodegenerative, immune, and lifestyle-related traits.

Those shared links are informative, but organ specificity is the central finding. A brain age gap overlaps with body-wide biology without replacing cardiovascular, liver, immune, or metabolic measures. Treating all age gaps as interchangeable would erase the very pattern that makes multi-organ work useful.

Why the genetics are useful for drug discovery

Fan Yi et al. used several genetic layers to rank possible targets. After their brain age-gap genome-wide association analysis, they combined cis-expression quantitative trait locus Mendelian randomization with colocalization and nominated 64 druggable genes. They then added protein quantitative trait locus Mendelian randomization and colocalization, which narrowed the list to seven stronger candidates: MAPT, TNFSF12, GZMB, SIRPB1, GNLY, NMB, and C1RL.

This is target prioritization, not a drug trial in disguise. Mendelian randomization can strengthen a causal hypothesis when its assumptions hold, and colocalization helps test whether genetic signals may arise from the same underlying variant. Neither method shows that giving a drug against one of these targets changes MRI age, dementia incidence, disability, mortality, or lifespan in people. The result is a better shortlist for experiments, not a prescription.

Multi-organ MRI research points in the same direction. A 2026 Nature Medicine study by Cao et al. developed MRI-based age gaps for the brain, heart, liver, adipose tissue, spleen, kidney, and pancreas in 313,645 people curated by the MULTI Consortium. The researchers linked the seven organ measures to 2,923 plasma proteins, 327 metabolites, and 6,477,810 common genetic variants, reporting 53 genome-wide significant organ-age-gap locus pairs and nine druggable genes for future anti-aging research. A related 2025 medRxiv preprint integrates similar MRI, molecular, and genetic data, but a preprint deserves less weight until peer review and independent replication.

Systemic risk travels with older-looking brains

Several observational studies connect an older-appearing brain with systemic physiology. A 2025 UK Biobank analysis by Feng et al. associated higher allostatic load with accelerated white-matter brain aging. Earlier UK Biobank work by Subramaniapillai et al. linked cardiometabolic risk with white-matter brain age in sex- and age-specific analyses.

Other studies widen the correlation map. In 17,770 chronic-brain-disorder-free UK Biobank participants aged 40 to 69, Yang Li et al. examined 249 baseline plasma metabolites and MRI-based brain age measured about nine years later; 77 metabolites were associated with brain age gap, with 55 also associated with predicted brain age. In 114 physically and cognitively healthy adults with overweight or obesity, Haas et al. found that poorer metabolic status was associated with older global and regional brain-age estimates. Longitudinal data existed for only 53 participants over three years, so that study cannot settle whether metabolic improvement changes brain age.

Diet data require the same restraint. A 2026 UK Biobank study included 21,473 neurologically healthy adults aged 40 to 70 and related a dietary inflammatory index, derived from repeated 24-hour recalls, to MRI brain age obtained roughly nine years after baseline. More pro-inflammatory diets were associated with older-looking brains. Association is not evidence that changing the diet will lower an age-gap score, much less prevent dementia through that score.

Proteomics offers a potentially more clinically relevant link, still without causal proof. Kou et al. reported that a plasma proteomics-based brain-aging signature predicted incident dementia risk in UK Biobank. Casanova et al. also associated plasma proteomics and age-related outcomes with brain age in a diverse cohort. These findings support biomarker research and risk stratification. They do not prove that the proteins cause dementia or that modifying them improves outcomes.

An older-looking brain is not necessarily a faster-aging brain

Model choice changes the meaning of the result. Whitman et al.'s DunedinPACNI study estimated a longitudinal Pace of Aging construct from a single brain image. That is an important advance because it aims to recover aging-rate information, rather than simply predicting chronological age. But it is not the same endpoint as the predicted-age-minus-chronological-age metric used in standard brain age-gap studies. A positive age gap and a faster pace of aging may overlap, yet the evidence does not permit treating them as identical.

What this means for biohackers

Brain and organ age gaps belong in research programs, where they can organize complex biology and help select pathways worth testing. The leading genetics studies point toward plausible targets and shared metabolic, immune, and neurodegenerative biology. They have not validated age gaps as surrogate endpoints for interventions that slow human aging.

The limitations are substantial: much of the evidence is cross-sectional or observational; key studies rely heavily on UK Biobank and predominantly European-ancestry samples; and age-prediction models can carry regression-to-the-mean and training-set bias. A one-year shift in brain age gap has no universally validated clinical meaning. Until randomized trials show that changing a treatment changes patient-important outcomes—and clarify whether an age-gap change mediates that benefit—a lower brain-age score remains an interesting measurement, not proof of rejuvenation.