Test script to to generate features on the PDBbind dataset, using DeepChem

deepchem_PDBbind_featurizer.py

# Ricardo Avila
# Script to use grid featurizer on the PDBbind dataset

import glob
import csv
import deepchem as dc
from deepchem.feat.rdkit_grid_featurizer import RdkitGridFeaturizer

receptors = glob.glob("./refined-set/*/*protein.pdb")
ligands = glob.glob("./refined-set/*/*ligand.sdf")
names = [n.split("/")[2] for n in glob.glob("./refined-set/*/*protein.pdb")]

GridFeaturizer = RdkitGridFeaturizer(voxel_width=16.0,
                                      feature_types=["all_combined"],
                                      ecfp_power=5,
                                      splif_power=5,
                                      flatten=True)

# Get features
for i in range(len(names)):
    try:
        print("Receptor", names[i])
        GridFeaturizer.featurize_complexes([ligands[i]], [receptors[i]])
    except:
        print("There was an error")

#for feature in features:
#    print(feature)
#    print("Number of features:", len(feature))