Python package for predicting single-cell CpG methylation states from DNA sequence and neighboring CpG sites using multi-task convolutional neural networks.

The DeepCpG git repository can be checked out into the current directory by executing the command

git clone https://github.com/cangermueller/deepcpg.git.

To install DeepCpG, execute python setup.py install in the root directory.

DeepCpG builds upon a fork of the deep learning framework Keras, which can be installed by running the setup.sh script in the tools/ directory, which will checkout and install the repository.

The examples/ directory contains example scripts for training a small deep neural network and predicting missing methylation states afterwards. The necessary data sets can be downloaded by running data.sh. A network can be trained by running train.sh, and methylation states for a test data set be predicted by running predict.sh.

• /deepcpg/: DeepCpG package
• /script/: Scripts for data generation, model training, and imputation
• /examples/: Example files for model training and imputation
• /data/: Example data files

• Christof Angermueller
• [email protected]
• https://cangermueller.com