Generating molecules with MolGAN

Deep graph generation is not well covered by PyTorch Geometric. Drug discovery is the main application of this subfield, which is why generative models can be found in specialized libraries. More specifically, there are two popular Python libraries for ML-based drug discovery: DeepChem and torchdrug. In this section, we will use DeepChem as it is more mature and directly implements MolGAN.

Let’s see how we can use it with DeepChem and tensorflow. The following procedure is based on DeepChem’s example:

1. We install DeepChem (https://deepchem.io), which requires the following libraries: tensorflow, joblib, NumPy, pandas, scikit-learn, SciPy, and rdkit:

   !pip install deepchem==2.7.1

2. Then, we import the required packages:

   import numpy as np
   import tensorflow as tf
   import pandas as pd
   from tensorflow import one_hot
   import deepchem as dc
   from deepchem.models.optimizers import ExponentialDecay
   from deepchem.models import BasicMolGANModel...