conda env create -f environment.yml
For docking, install QuickVina 2:
wget https://github.com/QVina/qvina/raw/master/bin/qvina2.1
chmod +x qvina2.1
MGLTools for preparing the receptor for docking (pdb -> pdbqt)
conda create -n mgltools -c bioconda mgltools
Starting a new training run:
python -u train.py --config <config>.yml
Resuming a previous run:
python -u train.py --config <config>.yml --resume <checkpoint>.ckpt
python test.py <checkpoint>.ckpt --test_dir <output_dir> --outdir <output_dir>
Under the Metrics folder, verify SA, QED, Div, Time, LogP, Lipinski
We follow the DiffSBDD method for verification. The verification method is as follows
First, convert all protein PDB files to PDBQT files using MGLTools
conda activate mgltools
cd analysis
python docking_py27.py <test_dir> <output_dir>
cd ..
conda deactivate
Then, compute QuickVina scores:
conda activate diff-fbdd
python analysis/docking.py --pdbqt_dir <docking_py27_outdir> --sdf_dir <test_outdir> --out_dir <qvina_outdir> --write_csv --write_dict