find-ligand-for-CB7

MCTS implementation branched off ChemTS (https://github.com/tsudalab/ChemTS) to find ligands for CB7's portal. The smiles dataset is located in data while the pretrained model and weights for smiles with at most 12 heavy atoms is located in RNN-model.

Set up the environment

Create a conda environment and activate it

conda create --name ligand-for-CB7 python=2.7

conda activate ligand-for-CB7

Install the rdkit

conda install -c rdkit rdkit

Install openbabel

conda install -c openbabel openbabel

Install requirements

python2 -m pip install -r requirements.txt

Edit the paths in xtb_binding.py

wdir = "PATH_TO/cap_design_for_CB7_XTB"

obabel_path = "PATH_TO/obabel"

ledock_path = "" # Ledock should be in the same folder as wdir, leave blank

xtb_path = "PATH_TO/xtb-6.3.3/bin/xtb"

Run the MCTS, i is an integer

cd cap_design_for_CB7_XTB

python2 mcts_ligand.py i

The results will be displayed to STDOUT then stored and compressed in the 'outputs' folder.

Train the MCTS

Edit the train_RNN.py file to set up the maximum smiles length (maxlen=31 by default). Then start training. See the ChemTS for training issues and how to train using GPUs.

cd train_RNN

python2 train_RNN.py