For the differential expression analysis, we curated two classes of gene sets to characterize the modules: Hypothesis-driven: This collection consisted of the hypothesis-driven sets dpreviously described with additional gene sets derived from previous cell type or region-specific studies or co-expression analyses.Cell type- or compartment-specific annotations:a) Cell type markers based on in situ hybridization in mouse brain tissue (abbreviated as ABA for Allen Brain Atlas123.b) Definite (10+ fold) enrichment for seven brain cell types, estimated based on FPKM for the given cell type vs. the average FPKM in the remaining types (abbreviated as Zhang124. For each cell type, only genes with FPKM > 1 were considered.c) Markers for different organelles and cellular compartments (markers of organelles, or MO)125–128d) Mitochondrial genes from the somatic vs. synaptic fraction of mouse cells (MitochondrialType) 129.Brain region-specific annotations: We used three categories of markers130:a) top 200 global marker genes for 22 large brain structures [globalMarker(top200)]. Genes were ranked based on fold change enrichment (expression in region vs. expression in rest of brain).b) top 200 local marker genes for 90 large brain structures [localMarker(top200)]. Same as a,
