Recently, I have been experimenting with a Graph Neural Network (GNN) and have attempted to train it on the FS-MOL dataset. However, I have used some additional preprocessing steps that were not included in the original FSMOL dataset.
It is crucial for neural network researchers to visualize the exact input that goes in the neural network. This ensures that the input data is correct and there are no errors in the preprocessing steps that may result in faulty data.
To depict a molecule, one common method is to use SMILES strings. First, we need to install the necessary libraries by running:
pip install rdkit ipywidgets nglview
To create a 3D depiction of a molecule, we can either download a preexisting 3D depiction of a molecule from sources like PubChem or use RDKit’s built-in module to calculate 3D coordinates of atoms in a molecule represented in the SMILES format.
Regardless of the method used, we require the nglview library to view 3D depictions of molecules in Jupyter Notebooks.
To demonstrate, let’s use RDKit to calculate the 3D coordinates of Remdesivir. The SMILES string for Remdesivir can be obtained from Wikipedia.
from rdkit import Chem
from rdkit.Chem import Draw, rdDepictor, AllChem
from rdkit.Geometry import Point2D
from IPython.display import display
import nglview as nv
smiles = 'CCC(COC(=O)[C@@H](NP(=O)(Oc1ccccc1)OC[C@H]1O[C@@]([C@@H]([C@@H]1O)O)(C#N)c1ccc2n1ncnc2N)C)CC' # The smiles string representing Remdesivir
mol = Chem.MolFromSmiles(smiles)
mol = Chem.AddHs(mol)
AllChem.EmbedMolecule(mol)
AllChem.MMFFOptimizeMolecule(mol) # Using the MMFFOptimization should result in a more accurate depiction
img = Draw.MolToImage(mol)
img.show()
The resulting image will be a 2D structure, as the Draw module can only depict 2D structures. However, we can improve this by creating an interactive 3D depiction of Remdesivir in Jupyter Notebooks by running:
view = nv.show_rdkit(mol)
view
The resulting ipywidget is an interactable 3D depiction of Remdesivir in Jupyter notebooks.