Disentangling Context-Specific Alzheimer's Disease Progression Signatures through an Explainable AI Framework

ADVISE disentangles Braak stage–linked transcriptional signals from background variation in human prefrontal cortex single-nucleus RNA-seq, prioritizing ROBO2 as a progression-associated candidate in excitatory neurons.

• Why it matters: Stage-dependent AD progression signatures are obscured by dominant cellular and biological variability, limiting interpretable identification of disease-specific genes and pathways.
• What they did: The explainable deep generative framework ADVISE learns separate latent representations for disease context versus background, then prioritizes genes and pathways across global, sex-specific, and cell-type-specific analyses.
• The result: In hiPSC-derived excitatory neurons, CRISPR activation of ROBO2 reduces the phosphorylated Tau-to-total Tau ratio, supporting ADVISE’s context-dependent candidate discovery.

AI summary of the abstract · may contain errors · not medical advice · read the paper.