Psychiatric Polygenic Scores Track Distinct Medical Comorbidity Patterns Across Biobanks

Psychiatric genetics is moving away from diagnosis labels built entirely on symptoms and toward risk architectures that may map more cleanly onto biology. In this large phenome-wide analysis, polygenic scores derived from five first-order psychopathology factors and one second-order hierarchical factor were tested against thousands of clinical outcomes and laboratory measurements across two healthcare biobanks. The result was not a single psychiatric signature, but a set of partially overlapping medical patterns that separated along cardiometabolic, inflammatory, digestive, respiratory, and immune pathways.

The scale matters. Investigators evaluated 108,182 individuals of European-like ancestry and 13,229 individuals of African-like ancestry, then meta-analyzed associations across 1,520 medical phenotypes and 123 laboratory metrics. This design allowed the study to move beyond a narrow psychiatric lens and ask whether genetic liability to broad psychopathology aligns with measurable differences in the rest of medicine. It does — and in several domains, the associations were strong enough to point toward biologically meaningful subgroups rather than one uniform psychiatric risk state.

One sentence captures the central finding: the transdiagnostic psychiatric factors did not behave as interchangeable constructs.

Medical Phenotypes Split Along Factor-Specific Genetic Liability Paths

The five first-order factors — compulsive, thought, neurodevelopmental, substance use, and internalizing disorders — showed robust links with as many as 143 conditions across 14 disease categories. The hierarchical factor reached 221 conditions across 15 categories, with odds ratios ranging from 0.75 to 1.94. Those are not trivial associations. They indicate that inherited liability to broad psychiatric phenotypes is reflected in medical comorbidity patterns that can be detected at the level of routine care records.

Circulatory, digestive, and respiratory diseases clustered most clearly with neurodevelopmental, internalizing, and substance use liabilities. Metabolic conditions moved in the opposite direction for the thought and compulsive factors, where the associations were negative, while the remaining three factors showed positive associations. In clinical terms, this means the same broad psychiatric architecture can point toward opposite directions of risk depending on which factor is under study. A genetic score associated with compulsive or thought-related liability does not simply mirror the score tied to internalizing or substance use risk.

The pattern is biologically plausible. Psychiatric dimensions are likely reflecting different developmental, autonomic, inflammatory, and behavioral pathways. Substance use liability often carries downstream effects on liver, respiratory, and cardiovascular health through exposure-related injury and shared reward circuitry. Internalizing liability may align with stress physiology and chronic illness burden. Neurodevelopmental liability may track with lifelong medical complexity, medication exposure, and altered health trajectories that accumulate across organ systems.

Laboratory Signals Point Toward Immunometabolic Subtypes

Laboratory data sharpened the picture. Lipid and glucose metabolism markers supported the phenotype associations, reinforcing the idea that psychiatric genetic liability can map onto cardiometabolic biology rather than appearing only in diagnostic coding. This matters because laboratory measures are closer to underlying physiology than many billing-derived disease labels. They offer a more direct window into the biology that may sit between inherited risk and clinical illness.

Immune markers were especially informative. C-reactive protein was specifically linked to the neurodevelopmental factor, while Complement C4 was specifically linked to the thought factor. Those two findings are small in number but large in implication. They suggest that different psychiatric dimensions may correspond to different immune-related mechanisms rather than a single generalized inflammatory state. CRP often signals systemic inflammation, whereas complement biology points toward a distinct arm of immune function with potential relevance to synaptic pruning, neurodevelopment, and broader host-defense regulation.

The laboratory results do more than echo the disease associations. They help explain why psychiatric genetic liability may appear in ordinary medical practice as diabetes risk, lipid abnormalities, or inflammatory activation rather than as overt psychiatric symptoms alone. For clinicians, that opens a practical route to stratification: patients with similar psychiatric diagnoses may carry very different immunometabolic profiles, and those profiles may shape treatment response, side-effect burden, or long-term physical health risk.

African-Like Ancestry Sample Reproduced the Broad Direction of Effects

A smaller African-like ancestry sample showed similar trends, lending support to the direction of the findings across ancestry groups. The sample size difference is important, and it limits the precision of subgroup estimates in the African-like ancestry cohort, but the replication of broad patterns argues against the entire signal being an artifact of one population. In a field where psychiatric genetics has too often depended on European ancestry data alone, any cross-ancestry consistency carries real weight.

Still, the study stays within the boundaries of association. Polygenic scores estimate inherited liability, not destiny, and phenome-wide signals do not prove causation. Medications, health care access, diagnostic practices, comorbid conditions, and other environmental forces can all shape how these associations appear in biobank records. Even so, the coherence of the medical and laboratory patterns makes the findings difficult to dismiss as noise.

Precision Psychiatry Now Has a Biomarker-Ready Rationale

The clearest clinical implication is not that psychiatric diagnosis should be replaced by genetics. It is that psychiatry may need to be organized around biologically informed subtypes that cut across traditional labels. The study directly supports ongoing efforts to define immuno-metabolic profiles and other transdiagnostic groupings that could improve patient stratification in clinical trials. If different genetic dimensions of psychopathology travel with different physical health risks, then trial design should account for those differences instead of assuming diagnostic uniformity.

This also has bedside relevance. A patient with elevated liability to one psychiatric factor may warrant closer attention to a different cluster of medical risks than a patient with elevated liability to another factor. The study does not provide a clinical decision rule, but it does provide a biological argument for why psychiatric care and general medicine should not be siloed. Mental illness risk appears to be entangled with cardiovascular, metabolic, respiratory, digestive, and immune biology in ways that are specific enough to matter.

What this work does not resolve is whether any of these associations can be converted into actionable prediction tools in routine practice. The current analysis establishes a broad phenotypic map, not a validated diagnostic assay. Future studies will need to test whether these genetic dimensions improve risk prediction beyond standard clinical variables, and whether the same patterns hold in more diverse populations with larger non-European ancestry representation. For now, the main advance is conceptual and biological: psychiatric polygenic scores are not only psychiatric. They reach into the rest of medicine in factor-specific ways that may eventually reshape how clinical subtypes are defined and studied.


Reporting by Dr. A. Manasa, MBBS for Medicine World Foundation, with additional data provided by medRxiv (Clinical Preprints).