juchem / lstm_molecular_gen

LSTM Molecular Gen

Trains LSTM with a dataset of SMILES molecular sequences, and uses the model to output new sequences.

Part of the code derived from:

• https://github.com/topazape/LSTM_Chem
• https://github.com/mattroconnor/deep_learning_coronavirus_cure

Running

The notebooks can be run under Jupyter Lab.

Dependencies

• matplotlib
• numpy
• openbabel
• pandas
• rdkit
• tensorflow

Sandboxed Container

A sandboxed container is provided with the complete environment needed to run the notebooks without installing additional software.

Build the docker image in Linux with make, make build or ./build.sh.

Run the docker container with make run (ensures image is built beforehand) or ./run.sh.

Open the lab's URL on your web browser (default: http://localhost:52019).

Configure Jupyter Lab with environment variables:

• NOTEBOOKS_DIR: notebooks root directory (default: repo root)
• LAB_ADDR: host address to bind the server to (default: 127.0.0.1)
• LAB_PORT: host port to bind the server to (default: 52019)

E.g.:

# run the lab on the default address and port
make run

# run the lab on http://127.0.0.2:51776
make run LAB_ADDR=127.0.0.2 LAB_PORT=51776