Nature Cover: 6.3M-Cell Atlas Reveals Largest Human Brain Gene Activity Map Across Aging and Disease

The PsychAD (Psychiatric and Neurodegenerative Disease) consortium published a series of coordinated studies in Nature and related journals, presenting the largest single-nucleus transcriptomic atlas of the human dorsolateral prefrontal cortex (DLPFC) to date. The dataset comprises 6.3 million nuclei from 1,494 postmortem brain donors aged 0–108 years, including healthy controls and eight brain disorders: Alzheimer’s disease, Parkinson’s disease, Lewy body dementia, vascular dementia, frontotemporal dementia, tauopathy, schizophrenia, and bipolar disorder. Over 30% of donors were of non-European ancestry, and ~21% had two or more co-occurring brain diseases, preserving age, genetic background, comorbidity, and clinical phenotype dimensions.

Disease Transcriptomic Landscape

Inter-individual variation explained ~10% of overall transcriptomic variance. Across diseases, a shared set of gene expression changes emerged, enriched for mRNA processing and protein localization. After removing these cross-disease effects, Alzheimer’s disease, Lewy body dementia, vascular dementia, and Parkinson’s disease showed stronger transcriptomic similarity. Within Alzheimer’s, disease severity correlated with neuronal loss and increased immune/vascular cells, while neuropsychiatric symptoms linked to deep-layer excitatory neuron alterations. The team constructed transcriptomic trajectories of disease progression.

Lifespan Molecular Dynamics in Neurotypical Brains

Using 1.3 million nuclei from 284 neurotypical donors (0–97 years), the study identified three broad phases of DLPFC molecular change: early development and remodeling, midlife relative stability, and late-life reactivation of specific molecular programs. A key inflection point occurred around age 24, after which most neuronal and glial subtype compositions stabilized, though some cell types continued changing. Late-life reactivated programs were concentrated in glial cells and associated with immune activation, stress response, and circadian rhythm reorganization. Most developmentally differentially expressed genes peaked around ages 12–13 before stabilizing.

Cell-Type-Specific Genetic Regulation

Genetic regulation analysis on 5.6 million DLPFC nuclei annotated into 8 major cell classes and 27 subclasses yielded 14,258 expression quantitative trait loci (eGenes). Of these, 981 showed cell-class-specific regulation, 857 showed subclass-specific regulation, 2,073 had effects varying along developmental trajectories, and 1,655 exhibited trans-regulatory effects. Colocalization with brain disease GWAS signals revealed cell-type-specific disease genes undetectable in bulk tissue analyses.

AI-Driven Phenotype Association with PASCode

A dedicated graph neural network framework, PASCode, scored cells for phenotype association across six diseases and neuropsychiatric phenotypes, identifying ~1.5 million phenotype-associated cells (PACs). This revealed Alzheimer’s pathology-linked microglial subsets and reactive astrocyte gene expression tied to cognitive resilience. PASCode re-groups cells by phenotype to uncover underlying genes and regulatory networks; it also distinguished glial changes in Alzheimer’s patients with depression from general Alzheimer’s inflammatory responses.

Individual-Level Functional Genomics with iBrainMap

Moving to the individual level, iBrainMap built personalized functional genomic graphs for over 1,900 brain samples, integrating cell-type interactions and gene regulatory networks via knowledge-guided graph neural networks. The model stratified donors by Alzheimer’s pathology and cognitive status into molecularly distinct subgroups and identified individual-specific cell types, genes, and regulatory networks. An interactive iBrainMap resource now provides open access to individual functional genomic graphs, models, and regulatory results.

Unified Reference Framework

By placing 1,494 donors and 6.3 million nuclei into a unified data space and re-reading the same brain region through disease contrast, normal aging, genetic regulation, Alzheimer’s phenotyping, and individual variation lenses, PsychAD created a large-scale human brain reference system that simultaneously accommodates age, disease, genetic background, cell type, and molecular state. All data, models, and analysis pipelines are publicly available, enabling joint comparison of questions previously studied in isolation.

Key Publications

Main atlas (Nature): https://www.nature.com/articles/s41586-025-09573-z Lifespan aging (Nature): https://www.nature.com/articles/s41586-026-10271-7 Genetic regulation (Nature Genetics): https://www.nature.com/articles/s41588-026-02733-5 PASCode AI analysis (Nature Medicine): https://www.nature.com/articles/s41591-025-04128-1 iBrainMap (Nature Communications):

https://www.nature.com/articles/s41467-026-72310-1