Can you provide the SMILES with labels for the ZINC dataset?

[DomInvivo]

Thank you for your work in providing a standard repository for graph benchmarking

Some applications might require using the SMILES to build different types of graphs than the one provided by the benchmarking platform. I know that the ZINC dataset come from the JT-VAE paper where the SMILES are provided. However, this paper is only a subset of the original dataset, and the train-val-test split is different.

I tried going from the DGLGraph back to SMILES, but it is not possible since I don't know which node label corresponds to which atom.

[vijaydwivedi75]

Hi @DomInvivo,
Thank you for the query.

Quick reply for:

D: I tried going from the DGLGraph back to SMILES, but it is not possible since I don't know which node label corresponds to which atom.

1. Please download the molecules_zinc_full.zip file from the link provided in this notebook. The zip contains library-agnostic ZINC data pickle files along with atom_dict.pickle and bond_dict.pickle
2. Read the atom and node dicts from the respective pickle files using this snippet, for knowing the node corresponds to which atom.

import pickle

class Dictionary:
    """
    word2idx and idx2word are mappings from words to idx and vice versa
    word2idx is a dictionary
    idx2word is a list
    word2num_occurence compute the number of times a given word has been added to the dictionary
    idx2num_occurence do the same, but with the index of the word rather than the word itself.
    """

    def __init__(self):
        self.word2idx = {}
        self.idx2word = []
        self.word2num_occurence = {}
        self.idx2num_occurence = []
        
data_folder = './'

with open(data_folder+"atom_dict.pickle","rb") as f:
    atom_dict=pickle.load(f)
with open(data_folder+"bond_dict.pickle","rb") as f:
    bond_dict=pickle.load(f)
    
print(atom_dict.word2idx)
print(bond_dict.word2idx)

OUTPUT:

{'C': 0, 'O': 1, 'N': 2, 'F': 3, 'C H1': 4, 'S': 5, 'Cl': 6, 'O -': 7, 'N H1 +': 8, 'Br': 9, 'N H3 +': 10, 'N H2 +': 11, 'N +': 12, 'N -': 13, 'S -': 14, 'I': 15, 'P': 16, 'O H1 +': 17, 'N H1 -': 18, 'O +': 19, 'S +': 20, 'P H1': 21, 'P H2': 22, 'C H2 -': 23, 'P +': 24, 'S H1 +': 25, 'C H1 -': 26, 'P H1 +': 27}
{'NONE': 0, 'SINGLE': 1, 'DOUBLE': 2, 'TRIPLE': 3}

--
Will get back soon for the SMILES data!

[DomInvivo]

Hey,
Thanks again for your answer! Did you manage to find the SMILES data of the ZINC dataset?

[vijaydwivedi75]

Hi @DomInvivo,

• The SMILES are here in the paper's repo, as you already mention.

• Corresponding to this, the full dataset (ZINC-full; 249K) is in this benchmarking repo. The order of molecules is same.

• The indices which corresponds to the molecules in the subset (ZINC; 12K) are in this folder.

Sorry for the late clarification.
Best Regards,
Vijay

[DomInvivo]

Thanks for your answer, this is helpful :)