The Allen Human Brain Atlas is an online multimodal atlas of the human brain that integrates anatomic and microarray-based gene expression information. Microarray sampling sites (~400-1000 sites per brain) were identified by expert anatomists using cytoarchitectural information from multiple histological stains. Sampling site delineations in the high resolution histological images were subsequently mapped into each individual's MR image space to provide 3-D anatomical context. All brains are also registered to MNI space to enable cross-individual comparisons.
RNA isolated from each sample area was hybridized to a custom Agilent 8x60k microarray chip to measure gene expression over the transcriptome. All least two different probes were available for 93% of genes. Probes were located on different exons as much as possible when multiple probes were available for a gene. For 60 genes, a set of tiling probes were probes was designed.
Each sampling site is associated to a Structure by expert anatomist using cytoarchitectural information from multiple histological stains. Structures are arranged in a hierarchical organization. Each structure has one parent and denotes a "part-of" relationship. Structures are assigned a color to visually emphasize their hierarchical positions in the brain.
Gene expression data for samples passing quality control were normalized first within a single brain using a cross-batch normalization algorithm called ComBat. In order to allow comparison across two or more brains, a cross-brain normalization was performed by aligning the mean 75th percentile expression values of all internal reference control samples of each brain to that of the first brain (H0351.2001). See the whitepaper for more details on microarray data generation and processing.
- All donors in the Product
- All sampling sites in the Product
- All microarray probes in the Product
- All microarray probes associated with gene prodynorhphin (PDYN)
- All samples associated with donor "H0351.2001" and the dentate gyrus
- Download an "raw" Aglient output file for one sample using the download-link(warning: large file)
Accessing the Microarray Data Service
A novel data service has been implemented to serve out normalized expression values and to provide on-the-fly differential and correlative searches over the entire microarray dataset.
Normalized expression values can be obtained by specifying a list of probes using the RMA connected services and pipes.
Example: Download expression values for all probes associated with PDYN
- Pipe:Find all probes associated with gene PDYN
- Connect pipe toservice::human_microarray_expressionto download expression values
The output of the service is two top level ordered arrays "samples" and "probes". For example:
Each sample contains information about:
- the Donor (id, name, age),
- the Sample (well id and (x,y,z) coordinate in the MR volume in millimeters),
- the associated Structure (id, name, acronym and color), and
- the associated top (coarse) level Structure (id, name, acronym and color).
Each probe contains information about:
- the Probe(id, name), and
- the Gene (id, acronym, name, entrez-id), along with
- a vector of normalized expression values in the same order as the "samples" array.
Differential search find probes that show the greatest difference between two sets (target and contrast) of user-defined structures. For each probe, a 2-sample t-test is performed followed by Benjamini and Hochberg false discovery rate correction. The null hypothesis is that the average expression level of samples in the contrast set of structures is less than the average expression level of samples in the target set of structures. Resulting p-values are sorted in ascending order. Search results can also be sorted by fold-change (log ratio of expression) in descending order.
Example: Differential search for genes with higher expression in thalamus than the cerebral cortex
- Pipe1: Set up contrast structure list by finding structure cerebral cortex in the Human Brain ontology
- Pipe2: Set up target structure list by finding structure thalamus in the Human Brain ontology
- Connect the two pipes toservice::human_microarray_differentialto perform the differential search
- Visualize the same search result in the Web application
Figure: Screenshot of top returns of a differential search for genes with higher expression in the thalamus than the cerebral cortex. Z-score heatmap shows enrichment (red) in the thalamus (structure color: light green) compare to other brain regions.
Usage of this service is demonstrated in the SPM example application.
Correlative search finds probes with similar expression profile to the selected seed probe over all samples within a user-specified structure. Pearson's correlation coefficient is computed for all probes and results ranked in descending order.
Example: Correlative search for probes with similar expression to PVALB probe CUST_11451_PI416261804 over the whole brain
- Pipe1: Set up the seed probe by finding PVALB probe CUST_11451_PI416261804
- Pipe2: Set up the domain structure list by finding structure Brain in the Human Brain ontology
- Connect the two pipes toservice::human_microarray_correlationto perform the correlation search
- Visualize the same search result in the Web application
Figure: Screenshot of top returns of a correlative search for probes with similar expression to a PVALB probe with expression values displayed as a z-score heatmap.
Magnetic Resonance Imaging
T1-weighted MPRAGE scans were acquired the postmortem brains using 3T Siemens Trio MR scanners (TI=900ms, TR=1900ms, TE=3.03ms, 9 degree flip angle, 1mm isotropic voxels). Scans were performed in cranio for some brains and ex cranio for others. See the white papers for more specific scan sequence details for each brain.
Figure: T1 MR scan for donor 'H0351.2002'
The T1, T2 and DTI (if available) volumetric data can be downloaded from the Web application or via the API.
All T1 images were registered to MNI space. FreeSurfer's affine registration was used for the in cranio scans. For ex cranio brains, the T1 was first rigidly aligned using FSL (Jenkinson, et. al, 2002) and then non-rigidly aligned using ANTS (Avants, et. al., 2011). The 3-D affine transform from a location in the MR volume to MNI space is encapsulated in the Alignment3d model.
See example code on how to transform each microarray sample to MNI space.