Exercise May Blunt the Genetic Risk of Depression, Landmark Study Finds

Depression runs in families, but it is not destiny. That familiar refrain has now received its most rigorous quantitative test to date. In a study published in Nature Mental Health, researchers drawn from the University of Texas at San Antonio, Vanderbilt Health, the University of Texas at Austin and Kansas State University analyzed more than 23,000 participants from the United States National Institutes of Health’s All of Us Research Program, combining whole-genome sequencing with years of wearable-device activity data. Their central question was deceptively simple: does physical activity buffer the impact of inherited genetic vulnerability to major depressive disorder? The answer, according to the data, depends on which face of depression you measure — and that distinction may reshape how scientists think about gene–environment interplay in mental health.

The scale and granularity of the dataset set this investigation apart from earlier attempts to probe the same question. The team studied 23,517 adults with a mean age of roughly 56 years, spanning 18 to 92 years old, of whom 67.2 percent were female. Crucially, physical activity was not assessed by questionnaire, the traditional weak link in behavioral research, but by consumer Fitbit devices worn over an average of 582 days — nearly a year and a half of continuous, objective monitoring. Activity was operationalized in two complementary ways: daily minutes of moderate-to-vigorous physical activity and daily step counts. Genetic liability was captured through polygenic risk scores, statistical summaries of thousands of common genetic variants that, in aggregate, index an individual’s inherited propensity toward depression.

Defining depression itself posed a methodological challenge that the authors addressed with unusual care. Rather than relying on a single data source, they constructed a dual-source phenotype combining electronic health record diagnoses with participants’ self-reported lifetime diagnoses of major depressive disorder. This triangulation matters because electronic health records are notoriously incomplete for mental health conditions, while self-report alone can be biased by recall and stigma. The researchers also examined symptom severity, derived from the health record, as a distinct outcome — a decision that proved pivotal. Most prior gene–environment studies of depression have focused almost exclusively on whether the disorder appears at all, treating it as a binary yes-or-no event and overlooking the continuous spectrum of symptom burden that clinicians actually encounter.

The first set of findings confirmed what the field has long suspected. Across every physical activity metric examined, a higher polygenic risk score was robustly associated with an increased likelihood of receiving a major depressive disorder diagnosis, while higher levels of physical activity were associated with lower risk. These main effects held after adjustment for demographic covariates and were consistent whether activity was measured in minutes of moderate-to-vigorous exertion or in steps. Yet when the researchers formally tested for interaction — that is, whether physical activity modified the relationship between genetic risk and the onset of diagnosis — the results were null. No significant polygenic risk score by physical activity interactions emerged for MDD onset, under any operationalization of activity or any definition of the diagnosis.

The picture changed dramatically when the outcome shifted from diagnosis to symptom severity. Here, both genetic risk and physical activity showed significant main effects, but the critical discovery was a consistent negative interaction: across all physical activity measures and across multiple definitions of severity, higher physical activity attenuated the association between genetic vulnerability and depressive symptom burden. In practical terms, individuals carrying high genetic risk who were more physically active tended to exhibit a lighter depressive symptom load than their genetically similar but sedentary counterparts. The interaction patterns proved robust across different model specifications, suggesting they were not artifacts of a particular statistical choice but a genuine feature of the data.

This dissociation — no interaction for onset, clear interaction for severity — is the study’s most intellectually provocative contribution. It implies that genetic and behavioral factors contribute largely independently to whether a person crosses the diagnostic threshold for depression, but that they interact in shaping how severe the illness becomes once it is expressed. Such a pattern aligns with a dimensional view of psychopathology, in which depression is understood as a continuum of symptom intensity rather than a discrete category. It also suggests that previous gene–environment studies, constrained to binary outcomes, may have been looking for buffering effects in the wrong place, or diluting them by collapsing a continuous phenotype into a crude dichotomy.

The biological plausibility of a protective effect of exercise on the brain is well developed in the literature. Physical activity has been linked to enhanced adult neurogenesis in the hippocampus, a brain region implicated in depression and memory; to increases in hippocampal volume with exercise training in humans; to modulation of neuroinflammatory pathways, given that depression has been characterized in part as a disorder involving chronic low-grade inflammation; and to improvements in stress-system regulation and psychological self-efficacy. Mendelian randomization studies have further suggested a bidirectional relationship, with physical activity causally reducing depression risk and depression, in turn, reducing activity levels. The new findings add a genomic dimension to this mechanistic picture, hinting that behavioral pathways may partially override or dampen the downstream expression of inherited risk.

The technical infrastructure behind the study deserves attention in its own right. Polygenic risk scores were computed using publicly available tools including PLINK 2.0 and SBayesRC, the latter leveraging functional genomic annotations to improve prediction accuracy. The All of Us program, which deliberately recruits a diverse American population and makes controlled-access data available through its Researcher Workbench, provided both the genomic sequences and the Fitbit streams; the analysis used the Control Tier dataset version 8. The authors were transparent about the limitations inherent in consumer wearables, whose validity relative to research-grade accelerometers has been evaluated extensively, and about the incompleteness of electronic health records, which is precisely why they combined record-based and self-reported diagnoses. Custom code for phenotype construction and statistical analysis was deposited publicly on Zenodo, an increasingly expected standard for computational reproducibility.

Cautious interpretation remains essential. This is an observational study, and even with nearly two years of device data, physical activity is not randomly assigned. People with higher genetic risk may move less because of subclinical symptoms, and reverse causation — depression reducing activity rather than activity preventing depression — can never be fully excluded in a cross-sectional analysis. The cohort, while large, consists of volunteers willing to wear trackers and share genomic data, a group that may differ systematically from the general population. Polygenic risk scores, moreover, currently explain only a modest fraction of depression’s heritability and perform unevenly across ancestries, a known limitation of genome-wide association data dominated by European-ancestry samples. The authors themselves frame physical activity as a potential modifier of genetic vulnerability, not a proven prescription.

Even with those caveats, the implications are striking. If physical activity preferentially dampens the expression of genetic risk in those who carry it, then exercise is not merely a general-purpose mood booster but a candidate tool for precision mental health — an intervention whose benefit may be greatest precisely for the individuals most biologically vulnerable. The findings also carry a public health message that transcends genomics: in a cohort of more than 23,000 people, measured over years with objective devices, more movement was consistently associated with less depression. As polygenic risk scoring inches toward clinical use, studies like this one suggest a future in which a genetic test might one day identify who stands to gain the most from a walking program — while reminding us that, for now, the simplest advice remains among the best supported: move more, and the mind may follow.

Subject of Research: Gene–environment interactions between polygenic risk for major depressive disorder and device-measured physical activity

Article Title: Associations and interactions between genetic vulnerability and physical activity with MDD in the All of Us Research Program

Article References: He, Q., Zhang, J., Beevers, C. G., & Brown, D. M. Y. (2026). Associations and interactions between genetic vulnerability and physical activity with MDD in the All of Us Research Program. Nature Mental Health. https://doi.org/10.1038/s44220-026-00738-x

Image Credits: AI Generated

DOI: 10.1038/s44220-026-00738-x

Keywords: major depressive disorder, polygenic risk score, physical activity, All of Us Research Program, gene-environment interaction, Fitbit, depression severity, Nature Mental Health, whole-genome sequencing, precision mental health, electronic health records, wearable devices