Polygenic Risk Splits Alzheimer’s and Psychiatric Presentations in Young-Onset Dementia
A real-world young-onset dementia cohort suggests that inherited risk for Alzheimer’s disease and psychiatric illness travels on separate tracks, even when plasma markers point toward neurodegeneration. In this prospective MiND Study analysis, polygenic risk scores for Alzheimer’s disease and for a cross-disorder psychiatric factor aligned with their expected diagnostic groups, yet remained independent of plasma NfL, GFAP, and p-tau217.
The finding matters because young-onset cognitive and behavioral syndromes often sit in the diagnostic gray zone where primary psychiatric disorders and neurodegenerative disease can look deceptively similar at presentation. In that setting, a biomarker of current tissue injury and a score that reflects lifelong inherited susceptibility are answering different clinical questions. One asks what is happening now. The other asks what the patient was carrying long before symptoms emerged.
Disease-Specific Signals Across 687 Participants
The cohort included 687 participants of European ancestry: 195 healthy controls, 283 with primary psychiatric disorders, 55 with Alzheimer’s disease, and 154 with non-Alzheimer neurodegenerative disease. Polygenic risk was calculated for Alzheimer’s disease, a cross-disorder psychiatric factor, frontotemporal dementia, Lewy body dementia, and Parkinson’s disease using PRS CS, with the APOE locus excluded. Plasma neurodegeneration markers were available for 332 participants, and the models were adjusted for age, sex, and APOE e4 dose.
In this dataset, the Alzheimer’s disease PRS was associated with Alzheimer’s disease relative to healthy controls, with an odds ratio of 1.69 per standard deviation and a 95% confidence interval of 1.17 to 2.45. The same score did not associate with primary psychiatric disorders, where the odds ratio was 0.95 with a 95% confidence interval of 0.77 to 1.16. The psychiatric polygenic signal showed the mirror image: the PSY PRS associated with primary psychiatric disorders, with an odds ratio of 1.29 and a 95% confidence interval of 1.05 to 1.58, but not with Alzheimer’s disease, where the odds ratio was 1.00 and the confidence interval was 0.72 to 1.40.
A single sentence captures the core result: genetic liability separated diagnoses, but it did not collapse into a single shared risk signature.
Plasma Biomarkers Reflected Injury, Not Polygenic Liability
No polygenic risk score associated with any plasma protein. That absence is clinically important because it implies a disconnect between inherited susceptibility and the measurable peripheral signature of ongoing neurodegeneration. Plasma biomarkers such as NfL, GFAP, and p-tau217 are closer to the present-state biology of axonal injury, astroglial activation, and tau pathology. Polygenic risk, by contrast, is upstream of symptoms and often upstream of biomarker change as well.
This biological separation offers a plausible explanation for why the two domains did not move together in the cohort. A person may carry elevated Alzheimer’s genetic risk without showing a corresponding plasma signature at the time of sampling, especially in earlier or diagnostically ambiguous stages. The same logic applies in the opposite direction: biomarker elevation can flag current neurodegenerative activity even when the inherited risk score is unremarkable.
Among participants in whom the question was primary psychiatric disorder versus non-Alzheimer neurodegenerative disease, NfL emerged as the strongest discriminator, with an odds ratio of 2.15 and a 95% confidence interval of 1.62 to 2.83. In this comparison, the Alzheimer’s PRS was also associated with non-Alzheimer neurodegenerative disease, with an odds ratio of 1.38 and a 95% confidence interval of 1.11 to 1.71, but the score did not materially improve discrimination when paired with NfL.
The area under the curve was 0.764 for NfL alone and 0.773 when the Alzheimer’s PRS was added. The gain was small, and in practical terms it suggests that a polygenic score may add context without meaningfully sharpening diagnostic separation in this specific analysis.
What the Results Mean for Real-World Diagnostic Workups
Young-onset dementia is not one disease category; it is a clinical sorting problem. Patients can present with apathy, disinhibition, memory complaints, cognitive slowing, or prominent psychiatric symptoms, and the initial differential may remain unstable until serial examination, imaging, and laboratory testing reveal the underlying process. These data support a two-layer framework: plasma biomarkers for active neurodegenerative biology and polygenic risk for inherited disease susceptibility.
In bedside terms, that framework could help explain why some patients with striking behavioral syndromes ultimately prove to have a psychiatric disorder while others carry a neurodegenerative diagnosis despite similar surface features. It also reinforces the idea that Alzheimer’s genetic risk and psychiatric genetic risk are not interchangeable proxies. Each seems to map onto its own diagnostic territory rather than serving as a generic index of brain disease.
The authors’ analysis also raises a caution about overextending polygenic scores derived from late-onset or population samples into young-onset presentations. Even when those scores preserve disease specificity, their value may be limited if they do not intersect with the biomarker biology that most directly tracks current tissue injury. In a cohort built around real-world referrals, that separation was visible rather than theoretical.
Why the Study Stays Preclinical and Why That Matters
This remains a preprint, and the diagnostic utility of combining polygenic risk with plasma biomarkers needs testing in larger, longitudinal pre-clinical cohorts. The available biomarker subset was smaller than the full cohort, and the analysis focused on European ancestry participants, which limits generalizability. The study also examined diagnostic association rather than prospective prediction of conversion, progression, or treatment response.
Even with those constraints, the pattern is unusually clean. Alzheimer’s disease polygenic risk aligned with Alzheimer’s diagnosis, psychiatric polygenic risk aligned with primary psychiatric illness, and neither score tracked with plasma markers of neurodegeneration. For clinicians facing a young patient whose presentation sits between psychiatry and neurology, that separation may prove useful—but not yet decisive.
The next step is longitudinal validation: whether these inherited risk signals add value before biomarkers rise, after symptoms begin, or only in narrow diagnostic corners where current tools still fail to separate disease from mimicry.