Marker Genes for Anatomical Regions in the Brain: Insights from the Allen Gene Expression Atlas

Abstract

Quantitative criteria are proposed to identify genes (and sets of genes) whose expression marks a specific brain region (or a set of brain regions). Gene-expression energies, obtained for thousands of mouse genes by numerization of in-situ hybridization images in the Allen Gene Expression Atlas, are used to test these methods in the mouse brain. Individual genes are ranked using integrals of their expression energies across brain regions. The ranking is generalized to sets of genes and the problem of optimal markers of a classical region receives a linear-algebraic solution. Moreover, the goodness of the fitting of the expression profile of a gene to the profile of a brain region is closely related to the co-expression of genes. The geometric interpretation of this fact leads to a quantitative criterion to detect markers of pairs of brain regions. Local properties of the gene-expression profiles are also used to detect genes that separate a given grain region from its environment.