-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathgraph_generator.py
379 lines (315 loc) · 20 KB
/
graph_generator.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
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
""" Graph Generator """
import pandas as pd
import torch
import sys
import dgl
import os
import networkx as nx
from rdkit import Chem
import pickle
import shutil
import zipfile
import arguments
from arguments import args
from utils_generator import create_descriptors, identify_functional_groups, \
CanonicalAtomFeaturizer, CanonicalBondFeaturizer, \
RandomScaffoldSplitter, RandomSplitter, RandomStratifiedSplitter_m,\
DatasetTox21, DatasetBBBP, DatasetBace, DatasetToxcast, \
DatasetClintox, DatasetSider, DatasetLipo, DatasetESOL, DatasetFreeSolv,\
DatasetHiv, DatasetMuv, DatasetQm7, DatasetQm8, \
DatasetPdbbind_r, DatasetPdbbind_c, DatasetPdbbind_f,\
fg_idx, fg_types_idx, convert_mol, make_df, fgs_connections_idx, \
graph_constructor, to_networkx, quotient_generator, splitted_data
current_dir = args.current_dir
dataset_names = {
"tox21" : "DatasetTox21",
"bbbp" : "DatasetBBBP",
"bace" : "DatasetBace",
"toxcast" : "DatasetToxcast",
"clintox" : "DatasetClintox",
"sider" : "DatasetSider",
"lipo": "DatasetLipo",
"esol" : "DatasetESOL",
"freesolv" : "DatasetFreeSolv",
"hiv" : "DatasetHiv",
"muv" : "DatasetMuv",
"qm7" : "DatasetQm7",
"qm8" : "DatasetQm8",
"pdbbind_r" : "DatasetPdbbind_r",
"pdbbind_c" : "DatasetPdbbind_c",
"pdbbind_f" : "DatasetPdbbind_f",
}
# generating global features
for name in args.gen_names_data:
name_global_csv = name + '_global_cdf_rdkit.csv'
name_global_zip = name + '_global_cdf_rdkit.zip'
saving_address = current_dir + 'data/global_features/' + name_global_zip
if not os.path.exists(saving_address):
print('Generating Global Features for', name)
raw_data_url = current_dir + 'data/raw/' + name + '.csv'
data = pd.read_csv(raw_data_url)
descriptors = create_descriptors(data, mols_column_name='smiles')
compression_opts = dict(method='zip', archive_name=name_global_csv)
os.makedirs(os.path.dirname(saving_address), exist_ok=True)
descriptors.to_csv(saving_address, index=False, compression=compression_opts)
# generating node features (for faster graph generation)
atom_featurizer = CanonicalAtomFeaturizer()
bond_featurizer = CanonicalBondFeaturizer()
for name in args.gen_names_data:
saving_address = current_dir + 'data/node_features/' + name + '_node_127_one_hot' + '.zip'
saving_address_pickle = current_dir + 'data/node_features/' + 'node_features.pickle'
if not os.path.exists(saving_address):
print('Generating Node Features for', name)
raw_data_url = current_dir + 'data/raw/' + name + '.csv'
data = pd.read_csv(raw_data_url)
node_features=[]
for i in range(len(data.smiles)):
node_features.append(atom_featurizer(Chem.MolFromSmiles(data.smiles[i]))['h'])
os.makedirs(os.path.dirname(saving_address), exist_ok=True)
with open(saving_address_pickle , 'wb') as handle:
pickle.dump(node_features, handle)
zf = zipfile.ZipFile(saving_address, 'w', zipfile.ZIP_DEFLATED)
zf.write(saving_address_pickle, 'node_features.pickle') #archname is necessary to remove the path once unpacked
zf.close()
os.remove(saving_address_pickle)
# generating splits
scaffold_splitter = RandomScaffoldSplitter()
random_splitter = RandomSplitter()
stratified_splitter = RandomStratifiedSplitter_m()
splitters = [scaffold_splitter, random_splitter, stratified_splitter]
type_indexs = {'scaffold' : 0, 'random' : 1, 'stratified' : 2}
# splitting
for name in args.gen_names_data:
raw_data_url = current_dir + 'data/raw/' + name + '.csv'
data = pd.read_csv(raw_data_url)
for split in args.splits:
for seed in args.generation_seeds:
saving_address = current_dir + 'data/splits/' + name + '/' + split + '_' + str(seed) + '/'
if not os.path.exists(saving_address + 'train_smiles') or not os.path.exists(saving_address + 'val_smiles') or not os.path.exists(saving_address + 'test_smiles'):
print('Generating', split, 'seed_', seed, 'split for', name)
splitted_sets = splitters[type_indexs[split]].split(data, seed=seed)
smiles_train = [data.smiles[i] for i in splitted_sets[0]]
smiles_val = [data.smiles[i] for i in splitted_sets[1]]
smiles_test = [data.smiles[i] for i in splitted_sets[2]]
os.makedirs(os.path.dirname(saving_address), exist_ok=True)
with open(saving_address + 'train_smiles', 'wb') as handle:
pickle.dump(smiles_train, handle, protocol=pickle.HIGHEST_PROTOCOL)
with open(saving_address + 'val_smiles', 'wb') as handle:
pickle.dump(smiles_val, handle, protocol=pickle.HIGHEST_PROTOCOL)
with open(saving_address + 'test_smiles', 'wb') as handle:
pickle.dump(smiles_test, handle, protocol=pickle.HIGHEST_PROTOCOL)
#generating graphs
for name_data in args.gen_names_data:
for name_model in arguments.generation_models_rows.keys():
for idx_row_graph_gen in arguments.generation_models_rows[name_model]:
## using generated_featurized_graphs for all seeds
generated_featurized_graphs = {}
are_graphs_generated = False
# specific settings for this scenario
dataset_name = dataset_names[name_data]
# """ Load csv of the specific graph generator"""
# save_csv = current_dir + "data/graph/"
# graph_gen_csv = pd.read_csv(save_csv + name_model+".csv")
# row_graph_gen = graph_gen_csv.iloc[idx_row_graph_gen]
if args.HQ_first_aggregation_op == 'sum':
print('Caution: Quotient by non-FGs with sum aggregation')
for split in args.splits:
print( '\n\n ========================= \n Generating', name_model, 'with row number', idx_row_graph_gen, '. Generating graphs', \
'for', split, 'splitting with seeds', args.generation_seeds, 'for',\
name_data, 'dataset is started! \n ========================= \n\n')
### Generation
# incorrect = False
list_gen_seeds = []
for seed in args.generation_seeds:
path_save= current_dir+"data/graph/"+name_data+"/"+split+"_"+str(seed)+"/"+arguments.name_final_zip
if not os.path.exists(path_save):
save_results_status = True
list_gen_seeds.append(seed)
# Print the current seed
print("Seed ", seed, " is started!")
name_node_feats_zip = name_data+ "_node"+arguments.name_node_feature+".zip"
name_global_csv = name_data+"_global_cdf_rdkit.csv"
name_global_zip = name_data+"_global_cdf_rdkit.zip"
"""Set Path"""
folder_data_temp = current_dir +"data/buffer/"
path_global_csv = folder_data_temp + name_global_csv
path_save_current_dir = folder_data_temp + arguments.name_final + "/"
path_save_temp = path_save_current_dir + split + "_" + str(seed)
path_save_0 = current_dir+"data/graph/"+name_data
path_save = current_dir+"data/graph/"+name_data+"/"+split+"_"+str(seed)+"/"
path_node_feats = current_dir + 'data/node_features/'
path_node_feats_zip = path_node_feats + name_node_feats_zip
path_smiles = current_dir + 'data/splits/' + name_data + "/" + split+"_"+str(seed)+"/"
path_data_csv = current_dir + 'data/raw/' + name_data + ".csv"
path_global_zip = current_dir + 'data/global_features/' + name_global_zip
#dataframe containing some details about the molecule
df = make_df(path_data_csv)
"""Hierarchical_Quotient"""
if name_model == "Hierarchical_Quotient":
def add_features(smiles, mol_dgl_graph):
# global incorrect
mol = Chem.MolFromSmiles(smiles)
# Add edge features
try:
mol_dgl_graph.edata["e"] = bond_featurizer(mol)["e"]
except:
# pass
mol_dgl_graph.edata["e"] = torch.zeros(mol_dgl_graph.num_edges(), 12)
### Adding a new column
mol_dgl_graph.edata["e"]=torch.cat((mol_dgl_graph.edata["e"], torch.ones(mol_dgl_graph.num_edges()).view(-1,1)), 1)
# Add node features
smiles_idx = dataset_smiles_series[dataset_smiles_series==smiles].index.values[0]
mol_dgl_graph.ndata["v"] = Node_features_loaded[smiles_idx]
### Adding a new column
mol_dgl_graph.ndata["v"]=torch.cat((mol_dgl_graph.ndata["v"], torch.ones(mol_dgl_graph.num_nodes()).view(-1,1)), 1)
#Quotient based on Carbons, FGs
mol_dgl_graph_q0 = quotient_generator(mol_dgl_graph, edge_condition_feature="edges_non_fgs", op=args.HQ_first_aggregation_op,\
another_edges_feature="edges_fgs").int()
if arguments.name_final == "Hierarchical_Quotient_type_False_Both_False_Uni_Vert_False_#quotient_2_#layers_1_127_one_hot":
mol_dgl_graph_q1 = quotient_generator(mol_dgl_graph_q0, edge_condition_feature="edges_fgs", op="sum").int()
mol_dgl_graph.ndata["qn2"]=torch.full((mol_dgl_graph.num_nodes(),1), 0).to(torch.int32)
mol_dgl_graph.edata["qe2"]=torch.full((mol_dgl_graph.num_edges(),1), 0).to(torch.float32)
####
mol_dgl_graph_q1.ndata["qn1"]=torch.full((mol_dgl_graph_q1.num_nodes(),1), -1).to(torch.int32)
mol_dgl_graph_q1.edata["qe1"]=torch.full((mol_dgl_graph_q1.num_edges(),1), -1).to(torch.float32)
####
mol_dgl_graph_q1.ndata["qn2"]= mol_dgl_graph_q1.ndata["qn2"].to(dtype=torch.int32)
mol_dgl_graph.ndata["qn1"]= mol_dgl_graph.ndata["qn1"].to(dtype=torch.int32)
mol_dgl_graph_q1.ndata["v"][:, -1] = 2
mol_dgl_graph_q1.edata["e"][:, -1] = 2
if arguments.name_final == "Hierarchical_Quotient_type_False_Both_False_Uni_Vert_False_#quotient_1_#layers_1_127_one_hot":
mol_dgl_graph_q0.ndata["v"][:, -1] = 1
mol_dgl_graph_q0.edata["e"][:, -1] = 1
return mol_dgl_graph_q0
elif arguments.name_final == "Hierarchical_Quotient_type_False_Both_False_Uni_Vert_False_#quotient_2_#layers_1_127_one_hot":
mol_dgl_graph_q1.ndata.pop('qn1')
mol_dgl_graph_q1.ndata.pop('qn2')
mol_dgl_graph_q1.edata.pop('qe1')
mol_dgl_graph_q1.edata.pop('qe2')
return mol_dgl_graph_q1
"""Replace global features NaN values with median (before making dataset)"""
if not os.path.exists(current_dir+name_global_csv):
shutil.unpack_archive(path_global_zip, folder_data_temp)
glob_csv = pd.read_csv(path_global_csv)
list_nan = []
for i in range(glob_csv.shape[1]):
if glob_csv.iloc[:,i].isnull().sum()>0:
list_nan.append(i)
print("list_nan_global_before: ", list_nan)
for i in list_nan:
glob_csv.iloc[:,i].fillna(glob_csv.iloc[:,i].median(), inplace=True)
list_nan = []
for i in range(glob_csv.shape[1]):
if glob_csv.iloc[:,i].isnull().sum()>0:
list_nan.append(i)
print("list_nan_global_after: ", list_nan)
glob_csv.to_csv(path_global_csv, index = False)
"""Load node features"""
if not os.path.exists(current_dir+"node_features.pickle"):
shutil.unpack_archive(path_node_feats_zip, folder_data_temp)
with open(folder_data_temp+"node_features.pickle", "rb") as handle:
Node_features_loaded = pickle.load(handle)
"""Make dataset"""
dataset = eval(dataset_name)(path_data_csv, path_global_csv)
# Define a pandas series from dataset smiles
dataset_smiles_series = dataset.smiles.squeeze()
# Train, validation, and test set split
train_set = splitted_data(path_smiles, dataset, dataset_smiles_series, string="train")
val_set = splitted_data(path_smiles, dataset, dataset_smiles_series, string="val")
test_set = splitted_data(path_smiles, dataset, dataset_smiles_series, string="test")
print(len(train_set), len(val_set), len(test_set))
"""Check if parameters of generator model are True or not"""
g = add_features(train_set[0][0], graph_constructor(df, train_set[0][0]))
# if incorrect:
# sys.exit("Seed "+str(seed)+" is exited, because the model settings are incorrect!")
# else:
# print("The model settings are correct!")
# nx.draw_networkx(g.to_networkx(), with_labels = True)
"""Prepare Training, Validation, and Test Sets"""
shutil.rmtree(folder_data_temp, ignore_errors=True)
#improving generation's speed with generating graphs only once for all seeds
#and splits in a (graph_model,row ID)
if not are_graphs_generated:
print('== \n Generating Featurized Graphs! \n ==')
for i in dataset_smiles_series:
generated_featurized_graphs[i] = add_features(i, graph_constructor(df, i))
are_graphs_generated = True
print(g, generated_featurized_graphs[train_set[0][0]])
"""Training Set"""
dgl_train=[]
smiles_train = []
labels_train= torch.empty(0)
masks_train = torch.empty(0)
globals_train = torch.empty(0)
counter = 0
for member in train_set:
bg = generated_featurized_graphs[member[0]]
dgl_train.append(bg)
smiles_train.append(member[0])
labels_train= torch.cat((labels_train, member[1]), dim=0)
masks_train= torch.cat((masks_train, member[2]), dim=0)
globals_train= torch.cat((globals_train, member[3]), dim=0)
counter+=1
print(counter)
label={"labels":labels_train, "masks":masks_train, "globals":globals_train}
new_path=path_save_temp+"_train.bin"
dgl.save_graphs(new_path, dgl_train, labels=label)
new_path=path_save_temp+"_smiles_train.pickle"
pickle_out = open(new_path,"wb")
pickle.dump(smiles_train, pickle_out)
pickle_out.close()
"""Validtion Set"""
dgl_val=[]
smiles_val = []
labels_val= torch.empty(0)
masks_val = torch.empty(0)
globals_val = torch.empty(0)
counter = 0
for member in val_set:
bg = generated_featurized_graphs[member[0]]
dgl_val.append(bg)
smiles_val.append(member[0])
labels_val= torch.cat((labels_val, member[1]), dim=0)
masks_val= torch.cat((masks_val, member[2]), dim=0)
globals_val= torch.cat((globals_val, member[3]), dim=0)
counter+=1
print(counter)
label={"labels":labels_val, "masks":masks_val, "globals":globals_val}
new_path=path_save_temp+"_val.bin"
dgl.save_graphs(new_path, dgl_val, labels=label)
new_path=path_save_temp+"_smiles_val.pickle"
pickle_out = open(new_path,"wb")
pickle.dump(smiles_val, pickle_out)
pickle_out.close()
"""Test Set"""
dgl_test=[]
smiles_test = []
labels_test= torch.empty(0)
masks_test = torch.empty(0)
globals_test = torch.empty(0)
counter = 0
for member in test_set:
bg = generated_featurized_graphs[member[0]]
dgl_test.append(bg)
smiles_test.append(member[0])
labels_test= torch.cat((labels_test, member[1]), dim=0)
masks_test= torch.cat((masks_test, member[2]), dim=0)
globals_test= torch.cat((globals_test, member[3]), dim=0)
counter+=1
print(counter)
label={"labels":labels_test, "masks":masks_test, "globals":globals_test}
new_path=path_save_temp+"_test.bin"
dgl.save_graphs(new_path, dgl_test, labels=label)
new_path=path_save_temp+"_smiles_test.pickle"
pickle_out = open(new_path,"wb")
pickle.dump(smiles_test, pickle_out)
pickle_out.close()
shutil.make_archive(path_save_current_dir, 'zip', path_save_current_dir)
os.makedirs(path_save, exist_ok=True)
b = os.path.join(path_save, arguments.name_final_zip)
shutil.copy(folder_data_temp+arguments.name_final_zip, b)
shutil.rmtree(folder_data_temp, ignore_errors=True)
print("Seed ", seed, " is finished!")
else:
list_gen_seeds.append(seed)
print("Seed ", seed, " was generated before!")