DeepBrain contains an example of using our enhancer models to score and understand any region in any genome for the cell types in our datasets.
git clone https://github.com/aertslab/DeepBrain.gitCreate and activate conda environment:
conda env create -f environment.yml
conda activate DeepBrainIf the installation fails for some reason, another option is to run the steps in the following script for a manual installation of all packages and the environment:
./install.shIf you are using a GPU (recomme 75A8 nded), and it is not found after installation, a potential fix may be to link an installed libcusolver.so.11 to the correct path:
#Define CUDA_INSTALL_PATH depending on where it is installed on the local machine
ln -s $CUDA_INSTALL_PATH/CUDA/11.3.1/lib64/libcusolver.so.11 $(python -c "import tensorflow.python as x; print(x.__path__[0])")/libcusolver.so.10The weights of the models are stored using Git Large File Storage (LFS). To download them, you will need to have installed Git LFS (https://git-lfs.com/). On Linux, you can install Git LFS with the following command if it was not installed yet:
sudo apt-get install git-lfs Then the following commands are required after installation to retrieve the model weights:
git lfs install
git lfs pullIf Git LFS does not work, you can also download the model weights from Zenodo: https://zenodo.org/records/10868679
Run the notebook DeepBrain_example.ipynb for example usage for predicting on genomic regions, getting contribution scores and calculating correlation between cell types. If you are running JupyterLab, you can make the environment visible by running:
ipython kernel install --user --name DeepBrain --display-name "DeepBrain"If the models or accompanying files are helpful for your research please cite the following publication:
Enhancer-driven cell type comparison reveals similarities between the mammalian and bird pallium
Nikolai Hecker*, Niklas Kempynck*, David Mauduit, Darina Abaffyová, Roel Vandepoel, Sam Dieltiens, Ioannis Sarropoulus, Carmen Bravo González-Blas, Elke Leysen, Rani Moors, Gert Hulselmans, Lynette Lim, Joris De Wit, Valerie Christiaens, Suresh Poovathingal, Stein Aerts