PASCode, a graph neural network framework for scoring phenotype-associated cells in population-scale single-cell data, identifies ~1.5 million Alzheimer's-phenotype-associated cells across the PsychAD cohort, prioritizing microglia, reactive astrocyte and excitatory/inhibitory neuron subpopulations linked to AD pathology, cognitive resilience, depression and disease progression.
Dreamlet is an open-source R package that uses a pseudobulk, precision-weighted linear mixed model approach to perform fast, scalable differential expression analysis across subjects in large single-cell/-nucleus transcriptomics datasets, demonstrated on a 1.4M-nuclei Alzheimer's disease brain dataset.
A lifespan single-nucleus atlas of over 1.3 million dorsolateral prefrontal cortex nuclei from 284 neurotypical donors reveals non-linear, cell-type-specific transcriptional trajectories, distinct neuronal and glial programs of early-life resilience and late-life vulnerability, and circadian reorganization with age, establishing a reference framework linking normative brain ageing to psychiatric and neurodegenerative disease risk.
iBrainMap, a knowledge-guided graph neural network framework, builds personalized functional genomic graphs from single-nucleus RNA-seq data across more than 1,900 brains to classify AD phenotypes, trace disease progression trajectories, prioritize donor-level cell-type genes and regulatory networks, and identify gene regulatory QTLs (grQTLs) linking genetic variants to donor-level regulatory changes.
A population-scale single-nucleus atlas of 1,494 donors and over 6.3 million DLPFC nuclei spanning eight brain disorders reveals shared cross-disease transcriptomic signatures, stronger genetic and transcriptomic concordance among Alzheimer's disease, DLBD, vascular dementia and Parkinson's disease, and cell-type-specific transcriptome trajectories capturing AD progression and associated neuropsychiatric symptoms.
A multi-ancestry single-nucleus atlas of genetic regulation across 5.6 million prefrontal cortex nuclei from 1,384 donors identifies cell-type-specific eQTLs and disease colocalizations for Alzheimer's disease, schizophrenia and other disorders that are undetectable in bulk tissue, including thousands of genes with dynamic developmental or trans-regulatory effects.
Single-nucleus transcriptomic imputation models (snTIMs) built from the multi-ancestry PsychAD cohort power a single-nucleus TWAS that uncovers thousands of cell-type-resolved gene–trait associations for neuropsychiatric and neurodegenerative disorders undetectable in bulk tissue, with pleiotropic, ancestry-conserved effects confirmed in the Million Veteran Program.
Profiling 832,505 prefrontal-cortex myeloid cells from 1,607 donors revealed a TREM2-dependent, neuroprotective GPNMB⁺ microglial state that expands with Alzheimer’s pathology, is enriched for polygenic AD risk and phagocytic activity, and is maintained by MITF.
Single-nucleus profiling of the adult human olfactory epithelium reveals that olfactory sensory neuron development shares transcriptional and regulatory programs with cortical neurogenesis, supporting its use as an accessible model for studying neurodevelopmental dysfunction in psychiatric disorders.
Taurine, a close analogue of tramiprosate, prevents and partially reverses ApoE4 aggregation and shifts ApoE4-associated molecular and cellular phenotypes toward ApoE3-like patterns, suggesting therapeutic potential for ApoE4-driven neurodegenerative risk.