Polygenic Risk Tracks With Distinct Laboratory Signatures Across Psychiatric Disorders
A laboratory-wide scan of polygenic scores is beginning to map inherited psychiatric risk onto measurable physiology. In a Penn Medicine Biobank analysis, investigators linked genetic liability for 25 health-related, psychiatric, and substance use phenotypes to routine clinical laboratory data and vital signs, using the electronic health record as a bridge between genomic risk and bedside phenotypes.
The central finding is not that psychiatric disorders share a single biomarker. They do not. Instead, the genetic architecture of these conditions appears to distribute across several recognizable physiological patterns, with some signals clustering around metabolism and inflammation, others around body weight, and a smaller set involving electrolytes and kidney-related measures. For disorders long considered biologically diffuse, that separation is clinically meaningful.
A Large EHR-Genomics Cohort Exposed Divergent Physiologic Patterns
The analysis included individuals of African ancestry, with AFR participants numbering 10,473, and individuals of European ancestry, with EUR participants numbering 35,131. The laboratory dataset was broad: 78 tests and six vital signs were run through a LabWAS pipeline, and linear regression models assessed each laboratory value against each polygenic score while adjusting for sex, age, and ten within-ancestry genetic principal components.
Four polygenic scores were significantly associated with at least one laboratory value in AFR participants, while 22 polygenic scores reached significance in EUR participants. The ancestry contrast is not a footnote. It signals how sample size, allele frequency architecture, and the limits of current discovery datasets can shape the visibility of genetic associations in clinical records. In practical terms, the EUR cohort yielded a far richer map of lab-linked physiology, while the AFR cohort produced fewer significant associations under the same analytic framework.
Metabolic and Inflammatory Signals Clustered in Anxiety, Depression, ADHD, and PTSD
In EUR participants, anxiety, attention-deficit/hyperactivity disorder, major depressive disorder, and post-traumatic stress disorder grouped into a shared physiological profile. The pattern included higher body mass index, higher glucose, and higher monocyte counts, along with lower high-density lipoprotein cholesterol and lower oxygen saturation.
This configuration points toward a metabolic and inflammatory axis rather than a purely neurochemical one. Elevated glucose and body mass index suggest cardiometabolic stress, while monocyte elevation is consistent with systemic immune activation. Lower HDL cholesterol fits the same direction of risk, and reduced oxygen saturation may reflect a broader physiologic burden that extends beyond mood symptoms alone. For clinicians, the pattern reinforces the idea that inherited liability for several common psychiatric disorders may intersect with metabolic health long before a formal psychiatric diagnosis is made.
The signal is clinically uncomfortable in the best possible way: it suggests that psychiatric genetics may be visible in standard laboratory data, not only in psychiatric assessments.
Weight-Related and Narrower Profiles Split Other Psychiatric Disorders Apart
A different set of psychiatric polygenic scores separated from the inflammatory-metabolic cluster. Anorexia nervosa, obsessive compulsive disorder, and schizophrenia were characterized by lower body mass index, indicating a shared direction for weight-related physiology even though the psychiatric phenotypes themselves are clinically distinct.
Bipolar disorder followed a more circumscribed pattern. Rather than the broad metabolic signature seen in anxiety, depression, ADHD, and PTSD, bipolar polygenic risk was linked to lower electrolytes and renal markers. The article does not extend the physiologic interpretation beyond those measures, but the narrowness of the signal is itself informative. Some disorders appear to project across multiple metabolic and immune markers, while others leave a smaller biochemical footprint that may reflect different underlying biology or different pathways of clinical ascertainment in the biobank.
Substance Use Polygenic Scores Followed Their Own Biochemical Routes
Substance use phenotypes did not collapse into a single group either. Tobacco use disorder and externalizing showed metabolic and inflammatory profiles similar to the cluster seen in several psychiatric disorders, including higher BMI, higher glucose, and higher monocytes, with lower HDL cholesterol and oxygen saturation.
Alcohol-related phenotypes were associated with lower red blood cell indices. Cannabis use disorder showed a more circumscribed profile, rather than the broader metabolic-inflamed pattern seen with tobacco use disorder and externalizing liability. Those differences matter because they imply that substance-related genetic risk is not physiologically interchangeable. Tobacco-related biology appears to overlap more strongly with systemic metabolic and inflammatory pathways, while alcohol-related risk mapped onto hematologic measures in this dataset.
What LabWAS Can and Cannot Say About Psychiatric Biomarkers
The value of this work lies in its use of routinely collected clinical data to create a phenotype layer between genotype and disease. Psychiatric genetics often struggles with translation because symptoms are heterogeneous and diagnoses are syndromic rather than mechanistic. A Laboratory-Wide Association Study can expose recurrent physiologic signatures that may be missed in case-control genetics alone, especially when the same disorder may present through different biologic routes in different patients.
Still, the design has limits. This is a cross-sectional association study using EHR laboratory data, so it cannot establish causality, temporal direction, or whether the lab abnormalities precede psychiatric illness, follow it, or arise from treatment, lifestyle, or comorbidity. The available evidence supports association, not prediction for individual patients. The data are also constrained by the biobank setting and by ancestry-specific sample size differences, which can affect which signals reach statistical significance.
Even with those limits, the study offers a concrete message for psychiatric biology. Polygenic risk for several disorders does not sit silently in the genome; in many patients, it appears to echo through standard bloodwork and vital signs. For future work, the critical task will be to determine which of these laboratory signatures are stable markers of inherited liability, which are downstream consequences of illness, and which could eventually help separate biologically distinct subtypes within broad psychiatric diagnoses.