Introduction

The brain is composed of diverse cell types including neurons (excitatory, inhibitory), astrocytes, oligodendrocytes, microglia, and endothelial cells. Bulk RNA-seq of brain tissue reflects a mixture of these cell types. Cell type deconvolution recovers cell type proportions from bulk expression data. Alzheimer's disease (AD) is the most common neurodegenerative disease, with genetic risk concentrated at loci including APOE, BIN1, CLU, and CR1.

Methods

Cell Type Deconvolution

CIBERSORT-style non-negative least squares regression using cell type signature matrices from single-cell RNA-seq references.

AD Risk Scoring

Polygenic risk score computed as weighted sum of GWAS risk alleles: PRS = Σ β_i × genotype_i.

Synaptic Network

Co-expression network constructed from synaptic gene set (SynGO database). Edges defined by |r|>0.5, FDR<0.05.

Neuroinflammation

Microglial activation score computed from DAM (disease-associated microglia) gene signature.

Results

Deconvolution r=0.998. AD risk p=2.26×10⁻⁵². AD risk score: 2.54 vs control: 0.58. Synaptic network: 500 edges.

Code Availability

https://github.com/BioTender-max/NeurogenomicsEngine

Key Results

• 200 brain samples (AD vs control)
• Deconvolution r=0.998
• AD risk p=2.26e-52
• AD score: 2.54 vs Ctrl: 0.58
• Synaptic edges: 500