In microarray studies of whole brain tissue, the measured expression level for any given gene depends on the extent to which it is expressed in different cell types and the representation of those cell types in the tissue sample. Therefore, it is commonly assumed in such studies that cellular heterogeneity precludes the recovery of cell type–specific information. Our analysis points to the opposite conclusion: cellular heterogeneity contributes in measurable and predictable ways to expression levels quantified by microarrays using messenger RNA extracted from whole brain tissue. To the best of our knowledge, this work is the first to provide evidence in support of this claim. However, this claim follows logically from two uncontroversial premises: cell types are distinguished by the genes that they express and the absolute quantity of each cell type will vary from sample to sample. Therefore, the genes that are most specifically and consistently expressed in the same cell type should appear highly correlated in microarray data derived from complex tissue homogenates. For a module consisting of such genes, the module eigengene can be interpreted as a
