Polygenic risk maps to distinct laboratory patterns across psychiatric disorders
A large LabWAS analysis has linked polygenic scores for psychiatric and substance use phenotypes to measurable physiology in routine clinical data, moving the field a step closer to biologic stratification of mental illness. Using Penn Medicine Biobank data, investigators examined 25 health-related, psychiatric, and substance use polygenic scores in two ancestry groups and paired those genetic risk measures with 78 laboratory tests and six vital signs. The study was built around a central clinical problem: psychiatric disorders carry heavy morbidity, yet they still lack robust diagnostic biomarkers grounded in everyday practice.
The signal was not diffuse. In individuals of European ancestry, 22 polygenic scores were associated with at least one laboratory value, compared with four in individuals of African ancestry. The asymmetry is clinically important because it reflects both the power of large-scale genetic association work and the reality that current genomic resources remain uneven across populations. Even so, the associations that did emerge point toward recurring physiologic patterns rather than isolated laboratory quirks.
One cluster linked anxiety, attention-deficit/hyperactivity disorder, major depressive disorder, and post-traumatic stress disorder to a shared metabolic and inflammatory profile. Higher body mass index, glucose, and monocytes tracked with lower high-density lipoprotein cholesterol and lower oxygen saturation. In bedside terms, this looks less like a single diagnostic biomarker and more like a systemic stress phenotype, where cardiometabolic burden, immune activation, and altered respiratory physiology travel alongside inherited psychiatric liability.
Metabolic and inflammatory signals dominated several psychiatric risk groups
The convergence of higher BMI, higher glucose, and monocyte elevation with lower HDL cholesterol is biologically plausible. Chronic low-grade inflammation has long been tied to depression and related disorders, while insulin resistance and dyslipidemia can shape both vascular health and brain function. Monocytes, in particular, may reflect innate immune activation that influences cytokine signaling and downstream neurobiologic pathways. Lower oxygen saturation adds another layer, suggesting that respiratory or autonomic influences may intersect with this metabolic-immune signature, though the study design cannot establish directionality.
A second pattern moved in the opposite direction for body habitus. Polygenic scores for anorexia nervosa, obsessive compulsive disorder, and schizophrenia were associated with lower BMI. The clustering is not identical across those conditions, yet the shared leaner profile may point to differences in energy balance, appetite regulation, or developmental biology that cut across diagnostic boundaries. In clinical practice, the finding is notable because it separates a body composition signal from the broader inflammatory-metabolic cluster seen in other psychiatric risk groups.
Bipolar disorder stood apart with a more circumscribed profile. Lower electrolytes and lower renal markers characterized that score rather than a broad metabolic-inflammatory footprint. This narrower pattern suggests a different physiologic route, one that may be less anchored to systemic inflammation and more related to renal handling, hydration status, or medication-related biology captured in the electronic record. The result does not identify a biomarker for bipolar disorder, but it does imply that not all psychiatric genetic risk resolves into the same peripheral laboratory signature.
Substance use polygenic scores showed their own physiologic groupings
The substance use findings tracked a parallel structure. Tobacco use disorder and externalizing phenotypes carried a metabolic and inflammatory profile similar to the psychiatric cluster described above. Alcohol-related phenotypes instead associated with lower red blood cell indices, while cannabis use disorder showed a more limited profile. These differences matter because they suggest that inherited liability for addictive behaviors may leave distinct biochemical footprints rather than a single nonspecific marker of risk.
Red blood cell indices can reflect nutritional status, marrow function, alcohol exposure, and chronic disease burden, all of which complicate interpretation in a biobank setting. Still, the association with alcohol-related scores fits a familiar clinical pattern in which chronic alcohol exposure often intersects with hematologic changes. Cannabis use disorder, by contrast, did not generate a broad laboratory signal in this analysis, which may reflect a narrower physiologic effect or limitations of available measures.
The study design used linear regression models adjusted for sex, age, and ten within-ancestry genetic principal components. That adjustment strategy aimed to reduce confounding from demographic structure and ancestry-related genetic background. Laboratory data were processed through a LabWAS pipeline, allowing the investigators to test associations systematically across a broad panel of common clinical tests and vital signs rather than relying on a single biomarker hypothesis.
What this means for psychiatric biomarker research
This work does not produce a diagnostic test for psychiatric illness. It does something more foundational: it shows that inherited risk for several psychiatric and substance use phenotypes can be read out in routine physiology, and that those readouts cluster in biologically coherent ways. For psychiatrists, internists, and translational researchers, the attraction is clear. If polygenic liability maps onto reproducible laboratory phenotypes, those phenotypes could help identify pathways worth targeting, refine disease subtyping, and guide mechanistic studies that extend beyond symptom checklists.
The ancestry imbalance in associations also points to a major limitation of current translational genomics. Four significant signals in African ancestry participants versus 22 in European ancestry participants is not just a statistical footnote; it is a reminder that discovery pipelines still depend heavily on the populations most represented in reference datasets. Any clinical use of this kind of biomarker work will require broader representation, or the resulting models will risk being precise for some patients and incomplete for others.
The preprint is explicitly not peer reviewed, and the findings should be treated as exploratory rather than ready for practice. Even so, the central biological message is difficult to ignore: psychiatric genetic risk is not confined to the brain. It appears to echo through inflammation, metabolism, hematology, and renal or electrolyte biology, leaving a peripheral trace that may eventually help define more precise disease subtypes. The next challenge is to determine which of these associations remain stable in independent cohorts and which are artifacts of a single health system biobank.