scAGG: Sample-level embedding and classification of Alzheimer's disease from single-nucleus data

Area of Science:

• Neuroscience
• Genomics
• Computational Biology

Background:

• Alzheimer's Disease (AD) pathogenesis requires identification of key cell types and genes.
• Single-cell RNA sequencing (scRNAseq) offers deep insights into AD at the cellular level.
• Classifying AD at the sample level from scRNAseq data presents unique challenges.

Purpose of the Study:

• To develop a method for sample-level classification of AD using scRNAseq data.
• To predict disease status from gene expression profiles of individual cells within a sample.
• To identify AD-associated cell subtypes and pathways for targeted therapeutic development.

Main Methods:

• Introduction of scAGG (single-cell AGGregation), a novel sample-level classification model.
• Utilizing a sample-level pooling mechanism to aggregate single-cell embeddings.
• Investigating the latent space learned by the model to correlate with disease severity.

Main Results:

• scAGG accurately classifies individuals with Alzheimer's Disease (AD) and healthy controls.
• The model's learned latent space reveals an ordering of cells corresponding to disease severity.
• Genes associated with disease severity are enriched in AD-linked pathways like cytokine signaling and apoptosis.

Conclusions:

• scAGG provides a robust method for phenotype classification from scRNAseq data, especially when cell-level annotations are absent.
• Cell- and sample-level severity scores generated by scAGG can identify AD-associated cell subtypes.
• This approach facilitates targeted drug development and personalized treatment strategies for AD.