The BANC (pronounced "bank") is the Brain And Nerve Cord, a GridTape-TEM dataset of an adult Drosophila melanogaster's entire central nervous system. This dataset was generated by Jasper Phelps and Minsu Kim in Wei-Chung Lee's lab at Harvard Medical School. A paper describing the dataset is currently in preparation.
This repository contains:
- A python package for interacting with the connectome data (see the folder
banc/
, and installation instructions below).- The
banc
package is a fork on thefanc
(Female Adult Nerve Cord) package, which can be found at the upstream repository. Note that to make it easier to synchronize new features inbanc
andfanc
packages, the folderbanc/
in this repo is a link to the folderfanc/
– but files in this repo's version of thefanc/
folder have been adapted for use with the BANC.
- The
- Other information about the dataset and about the community effort to proofread the connectome (see the wiki).
Have any questions? Please open an issue or contact Jasper Phelps ([email protected]).
As is always the case in python, consider making a virtual environment (using your preference of virtualenv/virtualenvwrapper or conda) before installing.
pip install banc
The code on GitHub will sometimes be slightly more up to date than the version on PyPI
pip install git+https://github.com/jasper-tms/BANC-fly-connectome.git
This is the best option if you want to make changes yourself to the code
cd ~/repos # or wherever you keep your repos
git clone https://github.com/jasper-tms/BANC-fly-connectome.git
cd BANC-fly-connectome
pip install -e .
Depending on your Python 3 version and your operating system, you may need to battle some bugs in order to get the pip install
commands above to succeed.
If you get something that looks like
.. ERROR:: Could not find a local HDF5 installation.
You may need to explicitly state where your local HDF5 headers and
library can be found by setting the ``HDF5_DIR`` environment
variable or by using the ``--hdf5`` command-line option.
and you're on a Mac, install brew
(https://brew.sh) if you haven't yet, then use brew
to install HDF5 with brew install hdf5
, then put HDF5_DIR=/opt/homebrew/opt/hdf5
in front of your pip install
command (e.g. HDF5_DIR=/opt/homebrew/opt/hdf5 pip install banc
).
If you get an error that contains
Error compiling Cython file:
...
Cython.Compiler.Errors.CompileError: tables/utilsextension.pyx
try to pip install
the latest version of tables
from GitHub by running HDF5_DIR=/opt/homebrew/opt/hdf5 pip install git+https://github.com/PyTables/PyTables
, or alternatively, use conda to install it (conda install tables
). After you get this package installed successfully, try installing banc
again.
Access to the latest reconstruction of BANC is restricted to authorized users. If you are a member of the BANC community (see Collaborative community on this repo's wiki) and have been granted access, you can generate an API key by visiting https://global.daf-apis.com/auth/api/v1/create_token and logging in with your BANC-authorized google account. Copy the key that is displayed, then run the following commands in python to save your key to the appropriate file:
import banc
banc.save_cave_credentials("THE API KEY YOU COPIED")
Alternatively, you can manually do what the command above accomplishes, which is to create a text file at ~/.cloudvolume/secrets/cave-secret.json
with these contents:
{
"token": "THE API KEY YOU COPIED",
"brain_and_nerve_cord": "THE API KEY YOU COPIED"
}
You can verify that your API key has been saved successfully by running:
import banc
client = banc.get_caveclient()
The mesh manipulation and coordinate transform code requires pytransformix
, which is itself a Python wrapper for Elastix. Therefore, Elastix must be installed and its lib and bin paths must be appended to the LD_LIBRARY_PATH
and PATH
environment variables. See pytransformix
documentation for specific instructions.
- First go through
fanc_python_package_examples.ipynb
- Then check out other notebooks in
example_notebooks/
- Finally you can browse the code, check out modules that have names that interest you, and read the docstrings.