-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathrepresentation.py
157 lines (133 loc) · 6.18 KB
/
representation.py
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
from tensorflow.python.keras import backend as K
from tensorflow.python.keras.models import model_from_yaml
from rdkit.Chem import AllChem
from rdkit.Chem import MACCSkeys
from optparse import OptionParser
import rdkit
import rdkit.Chem as Chem
import torch
import pandas as pd
import numpy as np
import sys
import subprocess
import os
sys.path.append('jtnn/')
from jtnn import *
sys.path.append('../')
class Representation:
def __init__(self, descriptor='jtvae'):
self.descriptor = descriptor
self.loaded_model = None
self.model = None
if self.descriptor == 'ecfp_autoencoder':
yaml_file = open('Models/autoencoder_optedilmemis.yaml', 'r')
loaded_model_yaml = yaml_file.read()
yaml_file.close()
self.loaded_model = model_from_yaml(loaded_model_yaml)
# load weights into new model
self.loaded_model.load_weights("Models/autoencoder_optedilmemis.h5")
elif self.descriptor == 'jtvae':
lg = rdkit.RDLogger.logger()
lg.setLevel(rdkit.RDLogger.CRITICAL)
# Jupyter notebookta hata verdiği için parser kapatıldı.
# parser = OptionParser()
# parser.add_option("-t", "--test", dest="test_path")
# parser.add_option("-v", "--vocab", dest="vocab_path")
# parser.add_option("-m", "--model", dest="model_path")
# parser.add_option("-w", "--hidden", dest="hidden_size", default=450)
# parser.add_option("-l", "--latent", dest="latent_size", default=56)
# parser.add_option("-d", "--depth", dest="depth", default=3)
# parser.add_option("-e", "--stereo", dest="stereo", default=1)
# opts, args = parser.parse_args()
vocab = [x.strip("\r\n ") for x in open("unique_canonical_train_vocab.txt")]
vocab = Vocab(vocab)
hidden_size = 450
latent_size = 56
depth = 3
stereo = True
model = JTNNVAE(vocab, hidden_size, latent_size, depth, stereo=stereo)
model.load_state_dict(torch.load("Models/model.iter-9-6000"))
# opts.model_path #MPNVAE-h450-L56-d3-beta0.001/model.iter-4
self.model = model.cuda()
def get_representation(self, smiles, descriptor, useChirality=False):
representation = [smiles]
mol = Chem.MolFromSmiles(smiles)
if mol is None:
return mol
if descriptor == "jtvae":
for i in self.jtvae_representation(smiles=smiles, mol=mol)[0]:
representation.append(i)
return np.asarray(representation)
elif descriptor == "ecfp_autoencoder":
for i in self.ecfp_representation(smiles=smiles, mol=mol):
representation.append(i)
return np.asarray(representation)
elif descriptor == "topological":
topological = Chem.RDKFingerprint(mol, fpSize=1024).ToBitString()
topological = np.fromstring(topological,'u1') - ord('0')
for i in topological:
representation.append(i)
return np.asarray(representation)
elif descriptor == "ecfp":
ecfp = AllChem.GetMorganFingerprintAsBitVect(mol, 3, nBits=1024,
useChirality=useChirality).ToBitString()
ecfp = np.fromstring(ecfp,'u1') - ord('0')
for i in ecfp:
representation.append(i)
return np.asarray(representation)
elif descriptor == "maccs":
maccs = MACCSkeys.GenMACCSKeys(mol).ToBitString()
maccs = np.fromstring(maccs,'u1') - ord('0')
for i in maccs:
representation.append(i)
return np.asarray(representation)
elif descriptor == "shed":
smi = smiles
mols = Chem.MolFromSmiles(smi)
hmols = Chem.AddHs(mols)
AllChem.EmbedMolecule(hmols,AllChem.ETKDG())
writer = Chem.SDWriter('smiles.sdf')
hmols.SetProp("_SMILES","%s"%smi)
writer.write(hmols)
writer.close()
shell = subprocess.call(["./shed.sh"])
data = pd.read_csv("myout1.csv" ,delimiter = "\t",sep = ":", header= None)
for i in data.columns.values:
if str(data.iloc[0][i]).find(":")!=-1:
data.iloc[0][i]=data.iloc[0][i][str(data.iloc[0][i]).find(":")+1:]
representation.append(data.iloc[0][i])
delete = subprocess.call(["./delete.sh"])
return np.asarray(representation)
elif descriptor == "cats2d":
smi = smiles
mols = Chem.MolFromSmiles(smi)
hmols = Chem.AddHs(mols)
AllChem.EmbedMolecule(hmols,AllChem.ETKDG())
writer = Chem.SDWriter('smiles.sdf')
hmols.SetProp("_SMILES","%s"%smi)
writer.write(hmols)
writer.close()
shell = subprocess.call(["./cats2d.sh"])
data = pd.read_csv("myout1.csv" ,delimiter = "\t",sep = ":", header= None)
for i in data.columns.values:
if str(data.iloc[0][i]).find(":")!=-1:
data.iloc[0][i]=data.iloc[0][i][str(data.iloc[0][i]).find(":")+1:]
representation.append(data.iloc[0][i])
subprocess.call(["./delete.sh"])
return np.asarray(representation)
def ecfp_representation(self, smiles, mol):
fp1 = AllChem.GetMorganFingerprintAsBitVect(mol, 3, nBits=1024)
a = np.asarray(fp1)
# with a Sequential model
get_3rd_layer_output = K.function([self.loaded_model.layers[0].input],
[self.loaded_model.layers[1].output])
layer_output = get_3rd_layer_output(np.asarray([a]))[0]
return layer_output[0]
def jtvae_representation(self, smiles, mol):
koku = pd.DataFrame(columns=list(range(56)))
smiles3D = Chem.MolToSmiles(mol, isomericSmiles=False)
dec_smiles = self.model.reconstruct(smiles3D, DataFrame=koku)
del smiles3D
del dec_smiles
torch.cuda.empty_cache()
return koku.values