File size: 3,152 Bytes
a48f0ae
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
# ---
# jupyter:
#   jupytext:
#     notebook_metadata_filter: -kernelspec
#     text_representation:
#       extension: .py
#       format_name: percent
#       format_version: '1.3'
#       jupytext_version: 1.14.1
# ---

# %% [markdown]
# ## General imports

# %%
import sys

sys.path.insert(
    0, "/"
)  # this depends on the notebook depth and must be adapted per notebook
# %%
import numpy as np
from compert.paths import DATA_DIR, EMBEDDING_DIR
from dgllife.utils import (
    CanonicalAtomFeaturizer,
    CanonicalBondFeaturizer,
    smiles_to_bigraph,
)

# %% [markdown]
# ## Load Smiles list

# %%
dataset_name = "lincs_trapnell"

# %%
import pandas as pd

smiles_df = pd.read_csv(EMBEDDING_DIR / f"{dataset_name}.smiles")
smiles_list = smiles_df["smiles"].values

# %%
print(f"Number of smiles strings: {len(smiles_list)}")

# %% [markdown]
# ## Featurizer functions

# %%
node_feats = CanonicalAtomFeaturizer(atom_data_field="h")
edge_feats = CanonicalBondFeaturizer(bond_data_field="h", self_loop=True)

# %% [markdown]
# ## Create graphs from smiles and featurizers

# %%
mol_graphs = []

for smiles in smiles_list:
    mol_graphs.append(
        smiles_to_bigraph(
            smiles=smiles,
            add_self_loop=True,
            node_featurizer=node_feats,
            edge_featurizer=edge_feats,
        )
    )

# %%
print(f"Number of molecular graphs: {len(mol_graphs)}")

# %% [markdown]
# ## Batch graphs

# %%
import dgl

mol_batch = dgl.batch(mol_graphs)

# %%
mol_batch

# %% [markdown]
# ## Load pretrained model

# %% [markdown]
# Choose a model form [here](https://lifesci.dgl.ai/api/model.pretrain.html)

# %%
model_name = "GCN_canonical_PCBA"
# model_name = 'MPNN_canonical_PCBA'
# model_name = 'AttentiveFP_canonical_PCBA'
# model_name = 'Weave_canonical_PCBA'
# model_name = 'GCN_Tox21'

# %%
from dgllife.model import load_pretrained

model = load_pretrained(model_name)

verbose = True
if verbose:
    print(model)

# %% [markdown]
# ## Predict with pretrained model

# %% [markdown]
# ### Take readout, just before prediction

# %%
model.eval()
# no edge features
prediction = model(mol_batch, mol_batch.ndata["h"])
# # with edge features
# prediction = model(mol_batch, mol_batch.ndata['h'], mol_batch.edata['h'])
print(f"Prediction has shape: {prediction.shape}")
prediction

# %% [markdown]
# ## Save

# %%
import pandas as pd

df = pd.DataFrame(
    data=prediction.detach().numpy(),
    index=smiles_list,
    columns=[f"latent_{i+1}" for i in range(prediction.size()[1])],
)

# %%
import os

fname = f"{model_name}_embedding_{dataset_name}.parquet"

directory = EMBEDDING_DIR / "dgl" / "data" / "embeddings"
if not directory.exists():
    os.makedirs(directory)
    print(f"Created folder: {directory}")

df.to_parquet(directory / fname)

# %% [markdown]
# Check that it worked

# %%
df = pd.read_parquet(directory / fname)
df

# %%
df.std()

# %% [markdown]
# ## Drawing molecules

# %%
from IPython.display import SVG
from rdkit import Chem
from rdkit.Chem import Draw

# %%
mols = [Chem.MolFromSmiles(s) for s in smiles_list[:14]]
Draw.MolsToGridImage(mols, molsPerRow=7, subImgSize=(180, 150))

# %%