Although cellular heterogeneity is a major organizing principle in the gene coexpression networks described in this study, we do not expect that every identified module will map unambiguously to a specific cell type in the brain for many reasons. For example, the definition of a cell type using currently available methods is inherently subjective1 and some modules are likely to constitute tightly integrated functional systems that span multiple cell types. Other modules capture variation in gene expression that is unrelated to cellular composition (for example, M1). Additionally, neuronal diversity in the brain produces a hierarchy of overlapping gene expression patterns, as evidenced by the large number of modules related to neuronal function identified in every network. Although some of these modules may reflect neuronal diversity (for example, M6), others probably represent transcriptional programs that are present in all neurons (for example, M16) or multiple neuronal subtypes (for example, M10 and M14).
