how to get molecular graph features

Belated I am interested in deepchem that is an open-source deep learning toolkit for drug discovery. Deep-chem supported many features for chemoinformatics.
And one of interested feature is calculation of molecular graphs. It is more primitive than hashed finger print. I tried to caluclate it.

Currently the toolkit supports only linux, so I installed deepchem via docker.
The installation was very easy.

iwatobipen$ docker pull deepchemio/deepchem
# wait a moment.... ;-)
iwatobipen$ docker run -i -t deepchemio/deepchem
iwatobipen$ pip install jupyter
# following code is not necessary.
iwatobipen$ apt-get install vim

That’s all.
Next, I worked in docker env.

import deepchem as dc
from deepchem.feat import graph_features
from rdkit import Chem
convmol=graph_features.ConvMolFeaturizer()
mol = Chem.MolFromSmiles('c1ccccc1')
# convmol needs list of molecules
fs = convmol.featurize( [mol] )
f = fs[ 0 ]
# check method
dir( f )
Out[41]:
[ .....
 'agglomerate_mols',
 'atom_features',
 'canon_adj_list',
 'deg_adj_lists',
 'deg_block_indices',
 'deg_id_list',
 'deg_list',
 'deg_slice',
 'deg_start',
 'get_adjacency_list',
 'get_atom_features',
 'get_atoms_with_deg',
 'get_deg_adjacency_lists',
 'get_deg_slice',
 'get_null_mol',
 'get_num_atoms',
 'get_num_atoms_with_deg',
 'max_deg',
 'membership',
 'min_deg',
 'n_atoms',
 'n_feat']

To get atom features, use ‘get_atom_features’
To get edge information, use ‘get_adjacency_list’

f.get_atom_features()
Out[42]:
array([[1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1],
       [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0,
        0, 0, 0, 0, 1, 0, 0, 0, 1]])
f.get_adjacency_list()
Out[43]: [[1, 5], [0, 2], [1, 3], [2, 4], [3, 5], [4, 0]]

The array of atom feature means, carbon atom, degree is 2, SP2, and aromatic as one hot vector.

Next step, I will try to build model by using molecular graph.