-
Notifications
You must be signed in to change notification settings - Fork 43
/
inference.py
238 lines (204 loc) · 13 KB
/
inference.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
"""
RUN INFERENCE
"""
import sys, os, subprocess, pickle, time, json, argparse
script_dir = os.path.dirname(os.path.realpath(__file__))
sys.path = sys.path + [script_dir+'/utils/'] + [script_dir+'/model/']
from sampler import *
sampler_map = {'default':SEQDIFF_sampler, 'cleavage_foldswitch':cleavage_foldswitch_SAMPLER}
def get_args():
"""
Parse command line args
"""
parser = argparse.ArgumentParser()
# design-related args
parser.add_argument('--pdb','-p',dest='pdb', default=None,
help='input protein')
parser.add_argument('--sequence',type=str, default=None,
help='input sequence to diffuse')
parser.add_argument('--trb', default=None, help='input trb file for partial diffusion')
parser.add_argument('--contigs', default='0', nargs='+',
help='Pieces of input protein to keep ')
parser.add_argument('--length',default=None,type=str,
help='Specify length, or length range, you want the outputs. e.g. 100 or 95-105')
parser.add_argument('--checkpoint', default=None,
help='Checkpoint to pretrained RFold module')
parser.add_argument('--inpaint_str', type=str, default=None, nargs='+',
help='Predict the structure at these residues. Similar mask (and window), but is '\
'specifically for structure.')
parser.add_argument('--inpaint_seq', type=str, default=None, nargs='+',
help='Predict the sequence at these residues. Similar mask (and window), but is '\
'specifically for sequence.')
parser.add_argument('--n_cycle', type=int, default=4,
help='Number of recycles through RFold at each step')
parser.add_argument('--tmpl_conf', type=str, default='1',
help='1D confidence value for template residues')
parser.add_argument('--num_designs', type=int, default=50,
help='Number of designs to make')
parser.add_argument('--start_num', type=int, default=0,
help='Number of first design to output')
parser.add_argument('--sampler',type=str, default='default',
help='Type of sampler to use')
# i/o args
parser.add_argument('--out', default='./seqdiff',
help='output directory and for files')
parser.add_argument('--dump_pdb', default=True, action='store_true',
help='Whether to dump pdb output')
parser.add_argument('--dump_trb', default=True, action='store_true',
help='Whether to dump trb files in output dir')
parser.add_argument('--dump_npz', default=False, action='store_true',
help='Whether to dump npz (disto/anglograms) files in output dir')
parser.add_argument('--dump_all', default=False, action='store_true',
help='If true, will dump all possible outputs to outdir')
parser.add_argument('--input_json', type=str, default=None,
help='JSON-formatted list of dictionaries, each containing command-line arguments for 1 '\
'design.')
parser.add_argument('--cautious', default=False, action='store_true',
help='If true, will not run a design if output file already exists.')
#diffusion args
parser.add_argument('--T', default=25, type=int,
help='Number of timesteps to use')
parser.add_argument('--F', default=1, type=int,
help='noise factor')
parser.add_argument('--save_all_steps', default=False, action='store_true',
help='Save individual steps during diffusion')
parser.add_argument('--save_best_plddt', default=True, action='store_true',
help='Save highest plddt structure only')
parser.add_argument('--save_seqs', default=False, action='store_true',
help='Save in and out seqs')
parser.add_argument('--argmax_seq', default=False, action='store_true',
help='Argmax seq after coming out of model')
parser.add_argument('--noise_schedule', default='sqrt',
help='Schedule type to add noise, default=cosine, could be [sqrt]')
parser.add_argument('--sampling_temp', default=1.0, type=float,
help='Temperature to sample input sequence to as a fraction of T, for partial diffusion')
parser.add_argument('--loop_design', default=False, action='store_true',
help='If this arg is passed the loop design checkpoint will be used')
parser.add_argument('--symmetry', type=int, default=1,
help='Integer specifying sequence repeat symmetry, e.g. 4 -> sequence composed of 4 identical repeats')
parser.add_argument('--symmetry_cap', default=0, type=int,
help='length for symmetry cap; assumes cap will be helix')
parser.add_argument('--predict_symmetric', default=False, action='store_true',
help='Predict explicit symmetrization after the last step')
parser.add_argument('--frac_seq_to_weight', default=0.0, type=float,
help='fraction of sequence to add AA weight bias too (will be randomly sampled)')
parser.add_argument('--add_weight_every_n', default=1, type=int,
help='frequency to add aa weight')
parser.add_argument('--aa_weights_json', default=None, type=str,
help='file path the JSON file of amino acid weighting to use during inference')
parser.add_argument('--one_weight_per_position', default=False, action='store_true',
help='only add weight to one aa type at each residue position (will randomly sample)')
parser.add_argument('--aa_weight', default=None, type=str,
help='weight string to use with --aa_spec for how to bias sequence')
parser.add_argument('--aa_spec', default=None, type=str,
help='how to bias sequence example XXXAXL where X is mask token')
parser.add_argument('--aa_composition', default=None, type=str,
help='aa composition specified by one letter aa code and fraction to represent in sequence ex. H0.2,K0.5')
parser.add_argument('--d_t1d', default=24, type=int,
help='t1d dimension that is compatible with specified checkpoint')
parser.add_argument('--hotspots', default=None, type=str,
help='specify hotspots to find i.e. B35,B44,B56')
parser.add_argument('--secondary_structure', default=None, type=str,
help='specified secondary structure string, H-helix, E-strand, L-loop, X-mask, i.e. XXXXXXHHHHHHXXXXLLLLXXXXXEEEEXXXXX')
parser.add_argument('--helix_bias', default=0.0, type=float,
help='percent of sequence to randomly bias toward helix')
parser.add_argument('--strand_bias', default=0.0, type=float,
help='percent of sequence to randomly bias toward strand')
parser.add_argument('--loop_bias', default=0.0, type=float,
help='percent of sequence to randomly bias toward loop')
parser.add_argument('--dssp_pdb', default=None, type=str,
help='input protein dssp')
parser.add_argument('--scheduled_str_cond', default=False, action='store_true',
help='if turned on will self condition on x fraction of the strcutre according to schedule (jake style)')
parser.add_argument('--struc_cond', default=False, action='store_true',
help='if turned on will struc condition on structure in sidneys style')
parser.add_argument('--struc_cond_sc', default=False, action='store_true',
help='if turned on will self condition on structure in sidneys style')
parser.add_argument('--softmax_seqout', default=False, action='store_true',
help='if turned on will softmax the Xo pred sequence before sampling next t')
parser.add_argument('--clamp_seqout', default=False, action='store_true',
help='if turned on will clamp the Xo pred sequence before sampling next t')
parser.add_argument('--no_clamp_seqout_after', default=False, action='store_true',
help='if turned on will clamp the Xo pred sequence before sampling next t')
parser.add_argument('--save_args', default=False, action='store_true',
help='will save the arguments used in a json file')
# potential args
parser.add_argument('--potential_scale', default=None, type=str,
help='scale at which to guid the sequence potential')
parser.add_argument('--potentials', default='',
help='list of potentials to use, must be paired with potenatial_scale e.g. aa_bias,solubility,charge')
parser.add_argument('--hydrophobic_score', default='0', type=float,
help='Set GRAVY score to guide sequence towards. Default == 0.0')
parser.add_argument('--hydrophobic_loss_type', default='complex', type=str,
help='type of loss to compute when using hydrophobicity potential')
parser.add_argument('--target_charge', default=0.0, type=float,
help='Set charge to guide sequence towards. Default == 0.0')
parser.add_argument('--target_pH', default=7.4, type=float,
help='Set pH to calculate charge at. Default == 7.4')
parser.add_argument('--charge_loss_type', default='complex', type=str,
help='type of loss to use when using charge potential')
parser.add_argument('--PSSM', default=None, type=str,
help='PSSM as csv')
# noise args
parser.add_argument('--sample_distribution', default="normal", type=str,
help='sample distribution for q_sample()')
parser.add_argument('--sample_distribution_gmm_means', default=[-1.0, 1.0], nargs='+',
help='sample distribution means for q_sample()')
parser.add_argument('--sample_distribution_gmm_variances', default=[1.0, 1.0], nargs='+',
help='sample distribution variances for q_sample()')
return parser.parse_args()
def main():
print(
'''
██████╗ ██████╗ ██████╗ ████████╗███████╗██╗███╗ ██╗
██╔══██╗██╔══██╗██╔═══██╗╚══██╔══╝██╔════╝██║████╗ ██║
██████╔╝██████╔╝██║ ██║ ██║ █████╗ ██║██╔██╗ ██║
██╔═══╝ ██╔══██╗██║ ██║ ██║ ██╔══╝ ██║██║╚██╗██║
██║ ██║ ██║╚██████╔╝ ██║ ███████╗██║██║ ╚████║
╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚══════╝╚═╝╚═╝ ╚═══╝
██████╗ ███████╗███╗ ██╗███████╗██████╗ █████╗ ████████╗ ██████╗ ██████╗
██╔════╝ ██╔════╝████╗ ██║██╔════╝██╔══██╗██╔══██╗╚══██╔══╝██╔═══██╗██╔══██╗
██║ ███╗█████╗ ██╔██╗ ██║█████╗ ██████╔╝███████║ ██║ ██║ ██║██████╔╝
██║ ██║██╔══╝ ██║╚██╗██║██╔══╝ ██╔══██╗██╔══██║ ██║ ██║ ██║██╔══██╗
╚██████╔╝███████╗██║ ╚████║███████╗██║ ██║██║ ██║ ██║ ╚██████╔╝██║ ██║
╚═════╝ ╚══════╝╚═╝ ╚═══╝╚══════╝╚═╝ ╚═╝╚═╝ ╚═╝ ╚═╝ ╚═════╝ ╚═╝ ╚═╝
'''
)
# parse args
args = get_args()
# chose sampler
chosen_sampler = sampler_map[args.sampler]
print(f'using sampler: {args.sampler}')
# initiate sampler class
S = chosen_sampler(vars(args))
# get JSON args
if args.input_json is not None:
with open(args.input_json) as f_json:
argdicts = json.load(f_json)
print(f'JSON args loaded {args.input_json}')
# wrap argdicts in a list if not inputed as one
if isinstance(argdicts,dict):
argdicts = [argdicts]
S.set_args(argdicts[0])
else:
# no json input, spoof list of argument dicts
argdicts = [{}]
# build model
S.model_init()
# diffuser init
S.diffuser_init()
for i_argdict, argdict in enumerate(argdicts):
if args.input_json is not None:
print(f'\nAdding argument dict {i_argdict} from input JSON ({len(argdicts)} total):')
### HERE IS WHERE ARGUMENTS SHOULD GET SET
S.set_args(argdict)
S.diffuser_init()
for i_des in range(S.args['start_num'], S.args['start_num']+S.args['num_designs']):
out_prefix = f'{args.out}_{i_des:06}'
if args.cautious and os.path.exists(out_prefix + '.pdb'):
print(f'CAUTIOUS MODE: Skipping design because output file '\
f'{out_prefix + ".pdb"} already exists.')
continue
S.generate_sample()
if __name__ == '__main__':
main()