-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathpipeline_gnn.py
203 lines (175 loc) · 8.22 KB
/
pipeline_gnn.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
import os
import sys
from utils import *
import sys
import warnings
import pdbreader
warnings.filterwarnings("ignore")
user_name = 'phucpht'
device = torch.device("cuda:4")
Seed_everything(seed = 42)
# ====== Global variables ======
folder_paths = sorted(glob(f"./drug_disease/*"))
idx = int(sys.argv[1])
if idx > len(folder_paths):
exit()
folder_path = folder_paths[idx]
if not os.path.exists(os.path.join(folder_path, 'preprocessed_data')):
exit()
protein_name = os.path.split(folder_path)[1]
dataset_path = 'datasets/drugbank.csv'
protein_path = os.path.join(folder_path, protein_name + '.pdb')
def prepare_protein():
remove_water_path = os.path.join(folder_path,'remove_water')
if not os.path.exists(remove_water_path):
os.makedirs(remove_water_path)
remove_water(protein_path, os.path.join(remove_water_path,f'{protein_name}.pdb'))
receptor_path = os.path.join(remove_water_path,f'{protein_name}.pdbqt')
if not os.path.exists(receptor_path):
convert_pdb_to_pdbqt(os.path.join(remove_water_path,f'{protein_name}.pdb'),receptor_path)
protein_remove_water = os.path.join(remove_water_path,f'{protein_name}.pdb')
try:
func = Vina_3d(
receptor_path,
[float(5), float(5), float(5)],
[80.0, 80.0, 80.0],
)
except:
print('Wrong protein')
exit()
return protein_remove_water
def pocket_position_to_coordinate(path_pocket):
pdb = pdbreader.read_pdb(path_pocket)
pocket_name = os.path.split(path_pocket)[1].replace('_atm.pdb', '')
residue_lst = " ".join(list(set([str(x) + ":" + str(a) for a, x in zip(pdb['ATOM']['resid'], pdb['ATOM']['chain'])])))
mean_pocket = np.mean(np.concatenate([np.array(pdb['ATOM'][dim_name]).reshape(-1,1) for dim_name in ['x','y','z']], axis = 1), axis = 0)
generation_path = os.path.join(folder_path, 'generation')
if not os.path.exists(generation_path):
os.makedirs(generation_path)
pocket_path = os.path.join(generation_path, f'{pocket_name}.sdf')
if not os.path.exists(pocket_path):
try:
cmd = f"python diffusion_generate/generate_ligands.py diffusion_generate/checkpoints/crossdocked_fullatom_cond.ckpt --pdbfile {receptor_path} --outfile {pocket_path} --resi_list {residue_lst} --n_samples 10 --device {device}"
os.system(cmd)
except:
print(f"Error {pocket_path}")
return mean_pocket, pocket_path
def mol_to_list_smiles(mol_file):
return sdf_to_smiles(mol_file)
def pocket_generation(protein_remove_water):
mean_pockets_x = []
mean_pockets_y = []
mean_pockets_z = []
smile_lists = []
key_idxs = []
idxs = 0
if len(glob(f'{folder_path}/remove_water/{protein_name}_out/pockets/*.pdb')) == 0:
cmd = f"fpocket -f {protein_remove_water}"
os.system(cmd)
for path_pocket in glob(f'{folder_path}/remove_water/{protein_name}_out/pockets/*.pdb'):
mean_pocket, pocket_path = pocket_position_to_coordinate(path_pocket)
try:
smile_list = mol_to_list_smiles(pocket_path)
except:
continue
for i in range(len(smile_list)):
mean_pockets_x.append(round(mean_pocket[0],2))
mean_pockets_y.append(round(mean_pocket[1],2))
mean_pockets_z.append(round(mean_pocket[2],2))
smile_lists.append(smile_list[i])
key_idxs.append(idxs)
data = {
'key_idx' : key_idxs,
'smiles' : smile_lists,
'x': mean_pockets_x,
'y': mean_pockets_y,
'z': mean_pockets_z,
}
idxs += 1
df = pd.DataFrame(data)
df.to_csv(os.path.join(os.path.split(pocket_path)[0], f'generation.csv'), index = False)
return os.path.join(os.path.split(pocket_path)[0], f'generation.csv')
def docking(generation_path, receptor_path):
outfile_path = os.path.join(folder_path, 'generation_docking', 'generation_docking.csv')
if os.path.exists(outfile_path):
return outfile_path
df = pd.read_csv(generation_path)
df['ba'] = np.zeros(len(df))
last_x, last_y, last_z = -99,-99,-99
for i in range(len(df)):
row = df.iloc[i]
smile = row['smiles']
x,y,z = row['x'], row['y'], row['z']
if (x != last_x or y != last_y or z != last_z):
func = Vina_3d(
receptor_path,
[float(x), float(y), float(z)],
[80.0, 80.0, 80.0],
)
last_x, last_y, last_z = x,y,z
smile = Chem.MolToSmiles(Chem.MolFromSmiles(smile), True)
LigPrepper.smiles2pdbqt(smile, labels=f"{protein_name}_{row['key_idx']}")
if not os.path.exists(os.path.join(folder_path, 'generation_docking')):
os.makedirs(os.path.join(folder_path, 'generation_docking'))
ba_generation = func(f"{protein_name}_{row['key_idx']}.pdbqt", os.path.join(folder_path, 'generation_docking',f"{row['key_idx']}.pdbqt"), n_poses = 5)
df.loc[i, 'ba'] = ba_generation
os.remove(f"{protein_name}_{row['key_idx']}.pdbqt")
df.to_csv(outfile_path, index = False)
return outfile_path
def preprocessed_data(path_generate_ligand):
out_file = os.path.join(folder_path, 'preprocessed_data', 'preprocessed_data.csv')
if not os.path.exists(os.path.join(folder_path, 'preprocessed_data')):
os.makedirs(os.path.join(folder_path, 'preprocessed_data'))
if os.path.exists(out_file):
return out_file
gconv = GConv(
input_dim=9, hidden_dim=64, activation=torch.nn.ReLU, num_layers=3
).to(device)
fc1 = FC(hidden_dim=64 * 3)
fc2 = FC(hidden_dim=64 * 3)
encoder_model = Encoder(encoder=gconv, local_fc=fc1, global_fc=fc2).to(device)
encoder_model.load_state_dict(torch.load(f"search_dgi/encoder_best.pt"))
encoder_model.eval()
generate = pd.read_csv(path_generate_ligand, encoding="utf-8")
dataset = pd.read_csv(dataset_path, delimiter=",")
docking_generation = []
for i in range(len(generate["smiles"])):
docking_generation.append(get_lig_graph(generate["smiles"][i]))
dataloader = DataLoader(docking_generation, batch_size=512, shuffle=False)
docking_generation_embed_lst = embed_data(encoder_model, dataloader, device)
datasets = []
for i in range(len(dataset["smiles"])):
datasets.append(get_lig_graph(dataset["smiles"][i]))
dataloader = DataLoader(datasets, batch_size=512, shuffle=False)
datasets_embed_lst = embed_data(encoder_model, dataloader, device)
docking_generation_embed = torch.stack(docking_generation_embed_lst).cpu().detach().numpy()
datasets_embed = torch.stack(datasets_embed_lst).cpu().detach().numpy()
gnns_sim = cosine_similarity(datasets_embed, docking_generation_embed)
potential_ligand_ba = list(generate["ba"])
gnns_x, gnns_y, gnns_z = list(generate["x"]), list(generate["y"]), list(generate["z"])
argmax_index = np.argmax(gnns_sim, axis=1)
gnns_array = np.hstack(
[
gnns_sim[np.arange(0, argmax_index.shape[0]), argmax_index].reshape(-1, 1),
np.array(potential_ligand_ba)[argmax_index].reshape(-1, 1),
np.array(gnns_x)[argmax_index].reshape(-1, 1),
np.array(gnns_y)[argmax_index].reshape(-1, 1),
np.array(gnns_z)[argmax_index].reshape(-1, 1),
]
)
dataset["gnns_score"] = gnns_array[:, 0]
dataset["gnns_ba"] = gnns_array[:, 1]
dataset["x"] = gnns_array[:, 2]
dataset["y"] = gnns_array[:, 3]
dataset["z"] = gnns_array[:, 4]
dataset["key_idx"] = dataset.index
data_sort = dataset.sort_values(by = ['gnns_ba','gnns_score'], ascending=[True, False])
data_sort['index'] = range(len(data_sort.sort_values(by = ['gnns_ba','gnns_score'])))
data_sort.to_csv(out_file, index=False)
return out_file
protein_remove_water = prepare_protein()
receptor_path = os.path.join(protein_remove_water,f'{protein_name}.pdbqt')
generation_path = pocket_generation(protein_remove_water)
path_generate_docking = docking(generation_path, receptor_path)
preprocessed_data_path = preprocessed_data(path_generate_docking)
print(f"Complete generate {preprocessed_data_path}")