Brain Gene Expression Tutorial

For up to date issues or problems with this database please see the wiki.

Brain Gene Expression is not a view into a single database, rather a view into three very large databases that all contain data about gene expression in brain regions.  The benefit of looking at these sources together is that they contain a very large amount of data, but the drawback is that although all data is presented uniformly, the data are not uniform in their annotation.

The following query will bring back data from all three sources, but any other gene or brain region query should bring back data from at least one of these sources. 

The BrainGeneExpression view can be found under “Nervous System Level” and “Multi-level.”  The first column contains the abbreviated database name, and the gene symbol and name are standard among these sources.  The links on the gene symbols lead to the three sources of data, where you can determine how the source data was generated.  Brain regions have some variability among sources in how they are defined. 

The Allen Brain Atlas (ABA) takes a well-defined stereotaxic coordinate indicating one of about 20 major, classical brain areas (hippocampus, cerebellum, and neocortex).  This means that many small regions or sub-nuclei will not be found in this source.  These data were not captured by ABA.

The Gensat project is annotated according to the Paxinos atlas (personal communication), with the explicit understanding that there are no true boundaries between the regions, making annotation difficult.  Gensat annotates to many more brain structures than ABA and sub-nuclei are also often mentioned in the notes.  However, it is important to keep in mind that the area discussed in the Gensat data as being neocortex may not overlap with the area labeled neocortex by ABA.

The MGI dataset incorporates a wide array of data sources with a great amount of variability in the annotations.  Although there is a nomenclature for mouse anatomy that is quite complete, data from individual papers is found in this source and this is taken verbatim from the authors, making it impossible to determine generally what regions were included in the labeled brain structures.  These data is intermingled with data from GenePaint and Deltagen which both have their own standards for brain region annotation.

The annotation of gene expression is more diverse, with a true/false annotation in MGI, qualitative annotation in Gensat and a numerical (1-100) annotation for ABA.  The NIF will not make any claims as to the meaning of the various annotations; we use this simply to say that there are data available in these sources that may match the investigators query.  Indeed, it should be noted that the MGI database is translating a very large set of qualitative annotations into true and false for the sake of presenting a simple view and the real data may not be well defined by the statement “true”.

Clicking on the Grm1 in the ABA row, leads to the Allen Brain Atlas database evidence record for the statement that Grm1 is expressed in the cerebellum.  Note, the cerebellum is the yellow column and it has the acronym CB. 

Clicking on the Grm1 in the Gensat rows, leads to the Gensat database evidence record for the statement that Grm1 is expressed in the cerebellum.  This page contains links to images of Grm1 expression.

Clicking on the Grm1 in the MGI rows, leads to the MGI database evidence records for the statement that Grm1 is expressed in the cerebellum.  Note, the variability of data records means that many of these landing pages will be quite different depending on which sorts of evidence is present for that particular gene.  It may be evidence extracted from supplementary material of a publication and it may be a record from the very large Deltagen knockout dataset.  In the example below, each of the columns is a piece of evidence from the Deltagen database where Grm1 was found to be expressed.  Searching through this site with patience will be needed to find the information how the data was obtained, but we have confidence that this is possible for all the data.