-
Notifications
You must be signed in to change notification settings - Fork 11
/
DeepAccNet.py
347 lines (293 loc) · 14.6 KB
/
DeepAccNet.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
#!/software/conda/envs/tensorflow/bin/python
import sys
import argparse
import os
from os import listdir
from os.path import isfile, isdir, join
import numpy as np
import pandas as pd
import multiprocessing
import torch
def main():
#####################
# Parsing arguments
#####################
parser = argparse.ArgumentParser(description="Error predictor network",
epilog="v0.0.1")
parser.add_argument("input",
action="store",
help="path to input folder or input pdb file")
parser.add_argument("output",
action="store", nargs=argparse.REMAINDER,
help="path to output (folder path, npz, or csv)")
parser.add_argument("--pdb",
"-pdb",
action="store_true",
default=False,
help="Running on a single pdb file instead of a folder (Default: False)")
parser.add_argument("--csv",
"-csv",
action="store_true",
default=False,
help="Writing results to a csv file (Default: False)")
parser.add_argument("--per_res_only",
"-pr",
action="store_true",
default=False,
help="Store per-residue accuracy only (Default: False)")
parser.add_argument("--leaveTempFile",
"-lt",
action="store_true",
default=False,
help="Leaving temporary files (Default: False)")
parser.add_argument("--process",
"-p", action="store",
type=int,
default=1,
help="Specifying # of cpus to use for featurization (Default: 1)")
parser.add_argument("--featurize",
"-f",
action="store_true",
default=False,
help="Running only the featurization part(Default: False)")
parser.add_argument("--reprocess",
"-r", action="store_true",
default=False,
help="Reprocessing all feature files (Default: False)")
parser.add_argument("--verbose",
"-v",
action="store_true",
default=False,
help="Activating verbose flag (Default: False)")
parser.add_argument("--bert",
"-bert",
action="store_true",
default=False,
help="Run with bert features. Use extractBert.py to generate them. (Default: False)")
parser.add_argument("--ensemble",
"-e",
action="store_true",
default=False,
help="Running with ensembling of 4 models. This adds 4x computational time with some overheads (Default: False)")
args = parser.parse_args()
################################
# Checking file availabilities #
################################
csvfilename = "result.csv"
# made outfolder an optional positinal argument. So check manually it's lenght and unpack the string
if len(args.output)>1:
print(f"Only one output folder can be specified, but got {args.output}", file=sys.stderr)
return -1
if len(args.output)==0:
args.output = ""
else:
args.output = args.output[0]
if args.input.endswith('.pdb'):
args.pdb = True
if args.output.endswith(".csv"):
args.csv = True
if not args.pdb:
if not isdir(args.input):
print("Input folder does not exist.", file=sys.stderr)
return -1
#default is input folder
if args.output == "":
args.output = args.input
else:
if not args.csv and not isdir(args.output):
if args.verbose: print("Creating output folder:", args.output)
os.mkdir(args.output)
# if csv, do it in place.
elif args.csv:
csvfilename = args.output
args.output = args.input
else:
if not isfile(args.input):
print("Input file does not exist.", file=sys.stderr)
return -1
#default is output name with extension changed to npz
if args.output == "":
args.output = os.path.splitext(args.input)[0]+".npz"
if not(".pdb" in args.input and ".npz" in args.output):
print("Input needs to be in .pdb format, and output needs to be in .npz format.", file=sys.stderr)
return -1
script_dir = os.path.dirname(__file__)
base = os.path.join(script_dir, "models/")
if not args.bert:
modelpath = join(base, "NatComm_standard")
else:
modelpath = join(base, "NatComm_bert")
# Eensemble is disabled right now.
if not isdir(modelpath):
print("Model checkpoint does not exist", file=sys.stderr)
return -1
##############################
# Importing larger libraries #
##############################
script_dir = os.path.dirname(__file__)
sys.path.insert(0, script_dir)
import deepAccNet as dan
num_process = 1
if args.process > 1:
num_process = args.process
#########################
# Getting samples names #
#########################
if not args.pdb:
samples = [i[:-4] for i in os.listdir(args.input) if isfile(args.input+"/"+i) and i[-4:] == ".pdb" and i[0]!="."]
ignored = [i[:-4] for i in os.listdir(args.input) if not(isfile(args.input+"/"+i) and i[-4:] == ".pdb" and i[0]!=".")]
if args.verbose:
print("# samples:", len(samples))
if len(ignored) > 0:
print("# files ignored:", len(ignored))
##############################
# Featurization happens here #
##############################
inputs = [join(args.input, s)+".pdb" for s in samples]
tmpoutputs = [join(args.output, s)+".features.npz" for s in samples]
if not args.reprocess:
arguments = [(inputs[i], tmpoutputs[i], args.verbose) for i in range(len(inputs)) if not isfile(tmpoutputs[i])]
already_processed = [(inputs[i], tmpoutputs[i], args.verbose) for i in range(len(inputs)) if isfile(tmpoutputs[i])]
if args.verbose:
print("Featurizing", len(arguments), "samples.", len(already_processed), "are already processed.")
else:
arguments = [(inputs[i], tmpoutputs[i], args.verbose) for i in range(len(inputs))]
already_processed = [(inputs[i], tmpoutputs[i], args.verbose) for i in range(len(inputs)) if isfile(tmpoutputs[i])]
if args.verbose:
print("Featurizing", len(arguments), "samples.", len(already_processed), "are re-processed.")
if num_process == 1:
for a in arguments:
dan.process(a)
else:
pool = multiprocessing.Pool(num_process)
out = pool.map(dan.process, arguments)
# Exit if only featurization is needed
if args.featurize:
return 0
if args.verbose: print("using", modelpath)
###########################
# Prediction happens here #
###########################
if args.bert:
samples = [s for s in samples if isfile(join(args.output, s+".features.npz")) and isfile(join(args.output, "bert_"+s+".npy"))]
else:
samples = [s for s in samples if isfile(join(args.output, s+".features.npz"))]
# Load pytorch model:
if args.ensemble:
modelnames = ["best.pkl", "second.pkl", "third.pkl", "fourth.pkl"]
else:
modelnames = ["best.pkl"]
result = {}
for modelname in modelnames:
model = dan.DeepAccNet(twobody_size = 49 if args.bert else 33)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
checkpoint = torch.load(join(modelpath, modelname), map_location=device)
model.load_state_dict(checkpoint["model_state_dict"])
model.to(device)
model.eval()
for s in samples:
try:
with torch.no_grad():
if args.verbose: print("Predicting for", s)
filename = join(args.output, s+".features.npz")
if args.bert:
bertname = join(args.output, "bert_"+s+".npy")
else:
bertname = ""
(idx, val), (f1d, bert), f2d, dmy = dan.getData(filename, bertpath = bertname)
f1d = torch.Tensor(f1d).to(device)
f2d = torch.Tensor(np.expand_dims(f2d.transpose(2,0,1), 0)).to(device)
idx = torch.Tensor(idx.astype(np.int32)).long().to(device)
val = torch.Tensor(val).to(device)
estogram, mask, lddt, dmy = model(idx, val, f1d, f2d)
t = result.get(s, [])
t.append(np.mean(lddt.cpu().detach().numpy()))
result[s] = t
if not args.csv:
if args.ensemble:
if args.per_res_only:
np.savez_compressed(join(args.output, s+"_"+modelname[:-4]+".npz"),
lddt = lddt.cpu().detach().numpy().astype(np.float16))
else:
np.savez_compressed(join(args.output, s+"_"+modelname[:-4]+".npz"),
lddt = lddt.cpu().detach().numpy().astype(np.float16),
estogram = estogram.cpu().detach().numpy().astype(np.float16),
mask = mask.cpu().detach().numpy().astype(np.float16))
else:
if args.per_res_only:
np.savez_compressed(join(args.output, s+".npz"),
lddt = lddt.cpu().detach().numpy().astype(np.float16))
else:
np.savez_compressed(join(args.output, s+".npz"),
lddt = lddt.cpu().detach().numpy().astype(np.float16),
estogram = estogram.cpu().detach().numpy().astype(np.float16),
mask = mask.cpu().detach().numpy().astype(np.float16))
except:
print("Failed to predict for", join(args.output, s+"_"+modelname[:-4]+".npz"))
if not args.csv:
if args.ensemble:
dan.merge(samples, args.output, verbose=args.verbose)
if not args.leaveTempFile:
dan.clean(samples,
args.output,
verbose=args.verbose,
ensemble=args.ensemble)
else:
# Take average of outputs
csvfile = open(csvfilename, "w")
csvfile.write("sample\tcb-lddt\n")
for s in samples:
line = "%s\t%.4f\n"%(s, np.mean(result[s]))
csvfile.write(line)
csvfile.close()
# Processing for single sample
else:
infilepath = args.input
outfilepath = args.output
infolder = "/".join(infilepath.split("/")[:-1])
insamplename = infilepath.split("/")[-1][:-4]
outfolder = "/".join(outfilepath.split("/")[:-1])
outsamplename = outfilepath.split("/")[-1][:-4]
feature_file_name = join(outfolder, outsamplename+".features.npz")
if args.verbose:
print("only working on a file:", outfolder, outsamplename)
# Process if file does not exists or reprocess flag is set
if (not isfile(feature_file_name)) or args.reprocess:
dan.process((join(infolder, insamplename+".pdb"),
feature_file_name,
args.verbose))
if isfile(feature_file_name):
# Load pytorch model:
#model = dan.DeepAccNet()
model = dan.DeepAccNet(twobody_size = 49 if args.bert else 33)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
#model.load_state_dict(torch.load("models/regular_rep1/weights.pkl"))
model.load_state_dict(torch.load(join(modelpath, "best.pkl"), map_location=device)['model_state_dict'])
model.to(device)
model.eval()
# Actual prediction
with torch.no_grad():
if args.verbose: print("Predicting for", outsamplename)
(idx, val), (f1d, bert), f2d, dmy = dan.getData(feature_file_name)
f1d = torch.Tensor(f1d).to(device)
f2d = torch.Tensor(np.expand_dims(f2d.transpose(2,0,1), 0)).to(device)
idx = torch.Tensor(idx.astype(np.int32)).long().to(device)
val = torch.Tensor(val).to(device)
estogram, mask, lddt, dmy = model(idx, val, f1d, f2d)
if args.per_res_only:
np.savez_compressed(outsamplename+".npz",
lddt = lddt.cpu().detach().numpy().astype(np.float16))
else:
np.savez_compressed(outsamplename+".npz",
lddt = lddt.cpu().detach().numpy().astype(np.float16),
estogram = estogram.cpu().detach().numpy().astype(np.float16),
mask = mask.cpu().detach().numpy().astype(np.float16))
if not args.leaveTempFile:
dan.clean([outsamplename],
outfolder,
verbose=args.verbose,
ensemble=False)
else:
print(f"Feature file does not exist: {feature_file_name}", file=sys.stderr)
if __name__== "__main__":
main()