Spatial brain maps

Spatial brain maps is a tool for viewing 3D gene expression in the Mouse brain. by fetching registration data shared via EBRAINS and ISH data shared via the Allen Institute, spatial brain maps reconstructs the 3D patterns of gene expression in a standardised coordinate space. Data can be reconstructed as either a 3D volume, or point cloud.

Usage

While this package is focused on creating 3D volumes, if you wish to search the maps and make region based queries use our public search interface hosted at https://neural-systems-at-uio.github.io/spatial_brain_maps/. This is integrated with the Siibra explorer atlas viewer so you can interactively search for genes which are up or down regulated in any given region, and explore those volumes of gene expression in 3D. The Python package by contrast allows you to generate new volumes for any given gene at different resolutions, in addition to creating point clouds data.

Installation

pip install spatial-brain-maps

you can view the PyPi package here

Quick CLI Usage

After installation a CLI command spatial_brain_maps is available.

1. Create a point cloud for an experiment ID.

experiment IDs can be found via the Allen Institutes mouse brain map portal. here we choose a resolution of 25 microns and only show values with an intesity greater than 30 (values are between 0 and 255).

spatial_brain_maps points --id 71717640 --mode expression --res 25 --cut 30
# Produces: 71717640_expression_cut30.json (MeshView compatible)

The above command produces a json file which you can view with MeshView

2. Create a point cloud for a gene (aggregate all experiments)

If you wish to aggregate all experiments for a particular gene we can provide --gene instead of --id

spatial_brain_maps points --gene Adora2a --mode expression --res 25 --cut 30

The above command for instance returns a json which when loaded in MeshView looks like this

3. Reconstruct a volume for an experiment ID and save to NIfTI

if we instead want a 3D volume we use the volume command like such.

spatial_brain_maps volume --id 71717640 --mode expression --res 25 --out-nifti outputs/exp71717640
# Produces: outputs/exp71717640.nii.gz

This produces a nifti file, which when viewed in software such as ITK Snap looks like this. Since we didn't choose to interpolate the data there are obvious gaps between the sections, this can be fixed by specifying --interpolate

4. Reconstruct an averaged gene expression volume (with interpolation)

in the same way as we are able to aggregate the data for the point clouds we can so again here. For the volumes we are also able to include the interpolate argument which fills the empty space between each section. Be careful as this is quite computationally intensive. If this is taking a long time you can instead choose a lower resolution.

spatial_brain_maps volume --gene Cnp --mode expression --res 25 --interpolate --out-nifti outputs/Cnp_mean

The above command produces this volume. Since the underlying experiments have each been interpolated before averaging the volume is smooth without obvious gaps.

Python API Examples

We also provide a Python interface which provides the same functionality

1. Reconstructing volumes

from spatial_brain_maps import gene_to_volume, write_nifti

# 1. Single experiment volume (returns a numpy array)
vol = gene_to_volume('Adora2a', resolution=25)

# 2. Save a volume to NIfTI 
write_nifti(vol, resolution=25, output_path="outputs/exp71717640")

This can be used alongside packages such as brainglobe-atlasapi for plotting and visualisation (see examples/plot_with_brainglobe.py)

2. Producing point clouds.

License

Distributed under the terms of the MIT License (see LICENSE).

Acknowledgements

• Allen Institute for Brain Science for raw ISH data, segmentations, and the Common Coordinate Framework.

---

Feel free to open issues or pull requests for feature requests, bug reports, or improvements.