-
Notifications
You must be signed in to change notification settings - Fork 5
/
predict.py
208 lines (179 loc) · 8.43 KB
/
predict.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
import sys
import os
import numpy as np
from sklearn.cluster import DBSCAN
sys.path.append("../")
from components.descriporConstruction.dataProcess import getProteinGrids
from components.descriporConstruction.gridConstruction import atomGridPositionToCoor
from components.descriporConstruction.fileOperation import readProtein
# get protein grids and block sampling by a step of 16
def prepareDataForPredict(config, mol2_file_name, pdbqt_file_name, feature_path, pdb, load_flag=False,
display_flag=False):
# protein grids construction
# print(feature_path)
if load_flag:
protein_channel1 = np.load(os.path.join(feature_path, pdb, "ligsite.npy"))
protein_channel2 = np.load(os.path.join(feature_path, pdb, "hbond.npy"))
protein_channel3 = np.load(os.path.join(feature_path, pdb, "vdw.npy"))
protein_channel4 = np.load(os.path.join(feature_path, pdb, "coulomb.npy"))
protein_grid = np.zeros((protein_channel1.shape[0], protein_channel1.shape[1], protein_channel1.shape[2], 4))
protein_grid[:, :, :, 0] = protein_channel1
protein_grid[:, :, :, 1] = protein_channel2
protein_grid[:, :, :, 2] = protein_channel3
protein_grid[:, :, :, 3] = protein_channel4
else:
protein_grid = getProteinGrids(config=config, mol2_file_name=mol2_file_name, pdbqt_file_name=pdbqt_file_name,
feature_path=None, pdb_name=pdb, buffer_size=8, resolution=1, train_flag=False,
display_flag=display_flag)
protein_grid = (2.0 * (np.arctan(protein_grid)) / np.pi)
# print(mol2_file_name)
# print(pdbqt_file_name)
protein = readProtein(config, mol2_file_name, pdbqt_file_name, pdb)
# for sampling
step_para = 4
temp_coor_list = [] # real cooridates for sampling blocks
temp_x_list = [] # sampling blocks
for i in range(0, protein_grid.shape[0] - 16, step_para):
for j in range(0, protein_grid.shape[1] - 16, step_para):
for k in range(0, protein_grid.shape[2] - 16, step_para):
temp_sample_block = protein_grid[i:i + 16, j:j + 16, k:k + 16, :]
temp_x_list.append(temp_sample_block)
temp_coor_list.append(
atomGridPositionToCoor([i + 8, j + 8, k + 8], protein.Mol2MinCoorNp, buffer_size=8,
resolution=1))
return temp_x_list, temp_coor_list
# DBSCAN for coordinates
def DBSCANCluster(probs, coors):
# print(probs)
# print(coors)
result_list_prob = []
result_list_center = []
step_para = 4.0
# DBSCAN parameters
eps, min_samples = (step_para + 1.0, 7)
# clustering
db = DBSCAN(eps=eps, min_samples=min_samples, metric='euclidean', algorithm='auto', leaf_size=30).fit(
np.array(coors))
# process with each clusters
labels = db.labels_
n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
# no effective clusters, use the class with -1 label
if n_clusters_ == 0:
cluster_coors = np.array(coors)[labels == -1]
cluster_probs = np.array(probs)[labels == -1]
temp_centers = np.mean(cluster_coors, axis=0)
# print(cluster_center)
temp_prob = np.mean(cluster_probs)
result_list_prob.append(temp_prob)
result_list_center.append(temp_centers)
# for each clusters
for i in range(n_clusters_):
cluster_coor = np.array(coors)[labels == i]
cluster_prob = np.array(probs)[labels == i]
temp_centers = np.mean(cluster_coor, axis=0) # center
# print(cluster_center)
temp_probs = np.mean(cluster_prob) # score
result_list_prob.append(temp_probs)
result_list_center.append(temp_centers)
# rank the results according to scores
result_list_top_index = np.argsort(np.array(result_list_prob))
result_list_prob = np.array(result_list_prob)[result_list_top_index][::-1]
result_list_center = np.array(result_list_center)[result_list_top_index][::-1]
return result_list_prob, result_list_center
# predict
def predict(config, mol2_file_name, pdbqt_file_name, model_path, feature_path, result_save_file, pdb, load_flag=False,
display_flag=False, top_pocket=3):
# get blocks and their coors
blocks_x, coors_x = prepareDataForPredict(config, mol2_file_name, pdbqt_file_name, feature_path, pdb,
load_flag=load_flag, display_flag=display_flag)
# model predict
from keras.models import load_model
model = load_model(model_path)
label_predict_x = model.predict(np.array(blocks_x))
# filter probs>=0.5
positive_coor_list = []
positive_prob_list = []
for l in range(label_predict_x.shape[0]):
if label_predict_x[l] >= 0.5:
positive_coor_list.append(coors_x[l])
positive_prob_list.append(label_predict_x[l])
# DBSCAN clustering
result_list_prob = [] # result for score
result_list_center = [] # result for center
# no blocks >0.5, choose the block with max pro. will not happen according to test.
if label_predict_x.shape[0] == 0:
result_list_top_index = np.argmax(label_predict_x)
result_list_prob.append(label_predict_x[result_list_top_index])
result_list_center.append(blocks_x[result_list_top_index])
else: # DBSCAN
result_list_prob, result_list_center = DBSCANCluster(positive_prob_list, positive_coor_list)
# for print and save
result_str = "result for " + pdb + "\r\n"
# top3 prediction output
pocket_num = result_list_prob.shape[0]
for i in range(min(top_pocket, pocket_num)):
pro = result_list_prob[i]
center = result_list_center[i]
result_str += "score: " + str(pro) + " predicted center: " + str(center) + "\r\n"
result_str += "number of all predicted pockets: " + str(pocket_num) + "\r\n"
print(result_str)
open(result_save_file, "w").writelines(result_str)
# top3 prediction
def predictTop3(config, mol2_file_name, pdbqt_file_name, model_path, result_save_file, pdb_name):
predict(config, mol2_file_name, pdbqt_file_name, model_path, config.feature_test_path, result_save_file, pdb_name,
load_flag=False,
display_flag=False, top_pocket=3)
# top5 prediction
def predictTop5(config, mol2_file_name, pdbqt_file_name, model_path, result_save_file, pdb_name):
predict(config, mol2_file_name, pdbqt_file_name, model_path, config.feature_test_path, result_save_file, pdb_name,
load_flag=False,
display_flag=False, top_pocket=5)
def printUsage():
print("python predict [protein.mol2] [protein.pdbqt] [3/5] [save_file]")
def main(config, argv):
try:
# print(argv)
if len(argv) != 4:
printUsage()
exit(0)
protein_mol2_file = argv[0]
protein_pdbqt_file = argv[1]
top_pocket = int(argv[2])
save_file = argv[3]
# print(protein_mol2_file,protein_pdbqt_file,top_pocket,save_file)
if ".mol2" not in protein_mol2_file and not os.path.exists(protein_mol2_file):
print("no mol2 file found.")
printUsage()
exit(0)
if ".pdbqt" not in protein_pdbqt_file and not os.path.exists(protein_pdbqt_file):
print("no pdbqt file found.")
printUsage()
exit(0)
if top_pocket != 3 and top_pocket != 5:
printUsage()
exit(0)
model_file = os.path.join(config.model_save_path, "model.h5")
if top_pocket == 3:
predictTop3(config, protein_mol2_file, protein_pdbqt_file, model_file, save_file, "protein")
else:
predictTop5(config, protein_mol2_file, protein_pdbqt_file, model_file, save_file, "protein")
except:
import traceback
traceback.print_exc()
print("An unexpected error occur.")
if __name__ == '__main__':
# gpu setting
# import tensorflow as tf
# from keras.backend.tensorflow_backend import set_session
#
# gpu_config = tf.ConfigProto()
# gpu_config.gpu_options.allow_growth = True
# set_session(tf.Session(config=gpu_config))
# os.environ["TF_CPP_MIN_LOG_LEVEL"] = "3"
#
from configure import Config
config = Config()
# # print(sys.argv)
# should prepare mol and pdbqt files for the predicted protein (open babel or autodock script)
# example: python predict.py example/1c6y_1/protein.mol2 example/1c6y_1/protein.pdbqt 3 ./results_example.txt
main(config, sys.argv[1:])