PsychAD maps 6.3 million brain-cell nuclei across eight disorders

The story

An international research consortium has released a population-scale molecular atlas of the human prefrontal cortex, built from more than 6.3 million individual cell nuclei donated by 1,494 people. Published in Nature, the PsychAD resource compares gene activity in neurotypical brains with patterns found across eight neurodegenerative and neuropsychiatric conditions. Its scale gives researchers a new way to ask which cellular responses are shared across diagnoses—and which remain specific to one disease.

The team used single-nucleus RNA sequencing on postmortem tissue from the dorsolateral prefrontal cortex, a region involved in executive function, working memory and emotional regulation. The cohort spans ages from infancy to 108 years, includes 723 male and 771 female donors, and draws tissue from the Mount Sinai NIH Neurobiobank, the National Institute of Mental Health Human Brain Collection Core and the Rush Alzheimer’s Disease Center. More than 30 percent of donors were classified as having non-European genetic ancestry, an important improvement over many earlier brain datasets, although it does not eliminate all representation gaps.

Researchers organized the nuclei into eight broad cell classes, 27 subclasses and 65 subtypes. They then compared neurotypical controls with donors affected by Alzheimer’s disease, diffuse Lewy body disease, vascular dementia, Parkinson’s disease, tauopathy, frontotemporal dementia, schizophrenia or bipolar disorder. That cross-condition design is central to the project: brain disorders often overlap clinically and biologically, while studies organized around a single diagnosis can miss molecular processes that cut across conventional disease labels.

The atlas found shared changes in basic cellular functions, including RNA processing and protein localization. After accounting for these broad signatures, the researchers reported stronger genetic and transcriptomic similarity among Alzheimer’s disease, diffuse Lewy body disease, vascular dementia and Parkinson’s disease. Neurodegenerative conditions generally showed a higher abundance of non-neuronal and vascular-associated cells, whereas the two psychiatric conditions in the comparison were more strongly associated with changes in neuronal populations. These are statistical patterns in donated tissue, not proof that any one cell type causes a disease.

Alzheimer’s disease received the deepest analysis. More severe disease was associated with lower neuronal abundance and increased immune and vascular cell populations. The researchers also modelled gene-expression trajectories across stages of pathology, pointing to cell-specific responses that differ between earlier and later disease. The result reinforces a view of Alzheimer’s as more than a disorder of neurons alone: immune activity, blood vessels and the interactions among multiple cell populations may all shape progression.

Two companion Nature papers extend the same resource. One charts cell-specific RNA expression across the human lifespan and reports dynamic remodelling during development, relative stability through much of adulthood and selective reactivation of developmental molecular programmes in later life. Another connects genetic risk for brain disorders to particular cell types using a single-nucleus transcriptome-wide association analysis. Together, the studies make the atlas useful not only for comparing diagnoses, but also for separating disease-linked signals from changes associated with ordinary ageing.

The work should not be read as a catalogue of ready-made drug targets. Postmortem tissue captures a final biological state and cannot by itself establish when a molecular change began or whether it drove disease rather than followed it. Diagnoses and comorbidities can overlap, and the analysis concentrates on one cortical region even though the conditions studied affect distributed brain systems. Candidate mechanisms will still need validation in independent cohorts, living-cell models and, eventually, carefully designed clinical research.

The project is unusually valuable as infrastructure. The consortium has made single-nucleus data available through the AD Knowledge Portal under controlled-access rules designed to protect donor privacy, provides an interactive CELLxGENE visualization and has released analysis code on GitHub. That combination should let outside teams test the published findings, compare their own cohorts and search the atlas for cell-specific patterns that might otherwise be hidden in bulk tissue measurements.

INNOVOX analysis: the important innovation is not a single headline association but the shift from small, disease-by-disease brain studies toward a reusable population-scale reference. A map this large can make weak hypotheses easier to reject and promising ones easier to prioritize. Its clinical value, however, will depend on whether patterns observed after death predict biology earlier in disease, whether they reproduce across populations and brain regions, and whether interventions aimed at those pathways change meaningful outcomes.

What to watch next is translation and replication. Researchers will need to test whether the shared signatures among Alzheimer’s, Lewy body disease, vascular dementia and Parkinson’s can improve disease classification or identify treatment strategies that cross diagnostic boundaries. More spatial data could show where the implicated cells sit and interact, while longitudinal biomarkers may connect postmortem molecular states to changes detectable during life. Expansion to additional brain regions and more globally representative cohorts will determine whether PsychAD becomes a durable reference rather than a very large snapshot.

INNOVOX analysis

PsychAD’s significance lies in turning a very large, carefully structured human-brain dataset into reusable research infrastructure. Its scale improves the ability to distinguish individual variation from disease-linked signals, but clinical impact requires independent replication, broader representation and evidence that the observed pathways matter earlier in disease.

What to watch

Watch for replication across populations and brain regions, spatial maps of the implicated cell interactions, links to biomarkers measurable during life, and experiments testing whether shared disease pathways yield actionable therapeutic targets.