Fig. 1

Data generation and analysis flow. a Schematic illustration of the 19 brain regions profiled in the current study. The numbered areas highlighted in yellow are the Brodmann (BM) areas, while the arrows indicate caudate nucleus (CD), nucleus accumbens (NAc), putamen (PT), amygdala (AMYG), and hippocampus (HIPP), respectively. b An overview of the analysis flow. RNA samples from 19 brain regions of 125 MSBB specimens were collected and profiled using Affymetrix Genechip microarrays. From the microarray RNA expression data, we first identified gene signatures associated with cognitive/neuropathological outcomes through differential expression and gene-trait correlation analyses. We tested enrichment of cell type-specific genes in the differentially expressed gene signatures and rank-ordered brain regions in relevance to AD by comprehensively comparing the number of gene signatures identified in each region for each trait. Next, we constructed a gene co-expression network for each brain region. Based on the network modules identified in individual brain regions, we constructed a meta-co-expression network to assess the correlation of networks between brain regions. Then we rank-ordered the co-expression network modules across all brain regions by multiple features. We evaluated the network module topology using gene perturbation signatures. Then, for top modules, we tested the replication of the network modules in an independent dataset from the Harvard brain bank. Later, we examined the cell type specificity and enrichment of genetic signal of the top ranked modules by using AD susceptibility genes and Aβ pathway genes. Finally, we explored regional selective vulnerability to the disease with two example pathways