forked from ChenShengsGitHub/AIxFuse_source
-
Notifications
You must be signed in to change notification settings - Fork 0
/
iter_gen.py
151 lines (141 loc) · 7.34 KB
/
iter_gen.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
import sys
import os
import pandas as pd
sys.path.append('utils')
sys.setrecursionlimit(3000)
import argparse
import warnings
import pickle
import numpy as np
import dual_MCTS
from ligpre import ligpre
from docking import docking
from train import train
from tqdm import trange
import time
warnings.filterwarnings('ignore')
task={'rorgt_dhodh':{'pdb_id1':'5NTP','pdb_id2':'6QU7','prec':'XP','device':'gpu','color':['#006400','#B22222']},'gsk3b_jnk3':{'pdb_id1':'6Y9S','pdb_id2':'4WHZ','prec':'SP','device':'cpu','color':['#87CEEB','#D2691E']}}
def iter_gen(args,iter):
target1,target2=args.task.split('_')
pdb_id1=task[args.task]['pdb_id1']
pdb_id2=task[args.task]['pdb_id2']
prec=task[args.task]['prec']
device=task[args.task]['device']
generated_dir=os.path.join(args.generated_dir,f'{target1}_{target2}')
cache_dir=os.path.join(args.cache_dir,f'{target1}_{target2}')
model_dir=os.path.join(args.model_dir,f'{target1}_{target2}')
iter_csv=os.path.join(generated_dir,f'gen_iter_{iter}.csv')
agent_dir=os.path.join(cache_dir,f'dual_MCTS_Agent_iter_{max(0,iter-1)}.pkl')
target1_pkl=os.path.join(cache_dir,f'core_info_{target1}.pkl')
target2_pkl=os.path.join(cache_dir,f'core_info_{target2}.pkl')
model_ckpt=os.path.join(model_dir,f'gen_iter_{max(0,iter-1)}.pt')
final_agent=os.path.join(cache_dir,f'dual_MCTS_Agent_iter_{iter}.pkl')
last_csv=os.path.join(generated_dir,f'util_iter_{max(0,iter-1)}.csv')
iter_mae=os.path.join(args.ligpre_dir,f'{target1}_{target2}_gen_iter_{iter}.maegz')
docking_dir='utils/docking'
dir_5ntp=os.path.join(docking_dir,f'{pdb_id1}_{prec}')
mae_5ntp=os.path.join(dir_5ntp,'docking_pv.maegz')
dir_6qu7=os.path.join(docking_dir,f'{pdb_id2}_{prec}')
mae_6qu7=os.path.join(dir_6qu7,'docking_pv.maegz')
grid_5ntp=os.path.join(args.grid_dir,f'{pdb_id1}.zip')
grid_6qu7=os.path.join(args.grid_dir,f'{pdb_id2}.zip')
docking_csv=os.path.join(generated_dir,f'gen_iter_{iter}_scores.csv')
next_csv=os.path.join(generated_dir,f'util_iter_{iter}.csv')
new_5ntp='/'.join(mae_5ntp.split('/')[:-1]+[f'gen_iter_{iter}.maegz'])
new_6qu7='/'.join(mae_6qu7.split('/')[:-1]+[f'gen_iter_{iter}.maegz'])
new_model=os.path.join(model_dir,f'gen_iter_{iter}.pt')
if not args.single_process or args.generate:
# 蒙特卡洛分子生成
if not os.path.exists(generated_dir):
os.makedirs(generated_dir)
if not os.path.exists(cache_dir):
os.makedirs(cache_dir)
if not os.path.exists(model_dir):
os.makedirs(model_dir)
# if not os.path.exists(iter_csv):
if iter<=1:
if os.path.exists(agent_dir):
with open(agent_dir,'rb') as r:
dual_MCTS_Agent=pickle.load(r)
else:
with open(target1_pkl,'rb') as r:
core_info_rorgt=pickle.load(r)
with open(target2_pkl,'rb') as r:
core_info_dhodh=pickle.load(r)
dual_MCTS_Agent=dual_MCTS.DualMCTS(core_info_rorgt,core_info_dhodh)
else:
with open(agent_dir,'rb') as r:
dual_MCTS_Agent=pickle.load(r)
if iter==0:
if not os.path.exists(iter_csv):
df=dual_MCTS_Agent.init_explore(args.gen_num)
df.to_csv(iter_csv,index=False)
with open(final_agent,'wb') as w:
pickle.dump(dual_MCTS_Agent,w)
else:
df_last=pd.read_csv(last_csv)
docking_result={}
for title,docking1,docking2 in zip(df_last['Title'],df_last[f'{pdb_id1}_{prec}'],df_last[f'{pdb_id2}_{prec}']):
docking_result[title]={f'{pdb_id1}_{prec}_neg':-docking1 if docking1 else 1,f'{pdb_id2}_{prec}_neg':-docking2 if docking2 else 1}
# if not os.path.exists(iter_csv):
final_gen=True if iter==5 else False
dual_MCTS_Agent.iter_explore(args.gen_num,model_path=model_ckpt,docking_result=docking_result,gen_csv=iter_csv,final_gen=final_gen)
with open(final_agent,'wb') as w:
pickle.dump(dual_MCTS_Agent,w)
if not args.single_process or args.ligpre:
# 生成分子的ligpre
ligpre(os.path.abspath(iter_csv),os.path.abspath(iter_mae),args.nchir,ligpre_path=args.ligpre_path,device=device)
for i in trange(100000):
time.sleep(10)
if os.path.exists(iter_mae):
break
if not args.single_process or args.docking:
# 生成分子的docking
docking(os.path.abspath(iter_mae),os.path.abspath(grid_5ntp),glide_path=args.glide_path,prec=prec,path=os.path.abspath(docking_dir),device=device)
time.sleep(10)
docking(os.path.abspath(iter_mae),os.path.abspath(grid_6qu7),glide_path=args.glide_path,prec=prec,path=os.path.abspath(docking_dir),device=device)
for i in trange(100000):
time.sleep(10)
if os.path.exists(mae_5ntp) and os.path.exists(mae_6qu7):
break
schrodinger_run=os.path.join(args.schrodinger_dir,'run')
os.system(f'{schrodinger_run} python3 utils/mae_to_csv.py '
f'--task get_scores --info_csv {iter_csv} '
f'--input_maes {mae_5ntp},{mae_6qu7} '
f'--output {docking_csv}')
df_next=pd.read_csv(docking_csv)
if iter==0:
df_next.to_csv(next_csv,index=False)
elif iter>=1:
df=pd.DataFrame({'SMILES':np.concatenate([df_last['SMILES'],df_next['SMILES']]),'Title':np.concatenate([df_last['Title'],df_next['Title']]),
f'{pdb_id1}_{prec}':np.concatenate([df_last[f'{pdb_id1}_{prec}'],df_next[f'{pdb_id1}_{prec}']]),
f'{pdb_id2}_{prec}':np.concatenate([df_last[f'{pdb_id2}_{prec}'],df_next[f'{pdb_id2}_{prec}']])})
df.to_csv(next_csv,index=False)
os.system(f'mv {mae_5ntp} {new_5ntp}')
os.system(f'mv {mae_6qu7} {new_6qu7}')
if not args.single_process or args.training:
# 生成分子的docking结果的训练
train(next_csv,new_model,pdb_id1,pdb_id2,prec)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--task',default='rorgt_dhodh')
parser.add_argument('--cache_dir',default='data/temp_data/')
parser.add_argument('--generated_dir',default='data/outputs/generated/')
parser.add_argument('--model_dir',default='data/models/dgl')
parser.add_argument('--gen_num',type=int,default=10000)
parser.add_argument('--iter',default='0,1,2,3,4')
parser.add_argument('--ligpre_dir',default='/public/home/chensheng/project/aixfuse2/data/outputs/ligpre/')
parser.add_argument('--nchir',default=32)
parser.add_argument('--grid_dir',default='data/inputs/target_structures/grids/')
parser.add_argument("--single_process", action="store_true")
parser.add_argument("--generate", action="store_true")
parser.add_argument("--ligpre", action="store_true")
parser.add_argument("--docking", action="store_true")
parser.add_argument("--training", action="store_true")
parser.add_argument('--schrodinger_dir', required=True)
parser.add_argument('--ligpre_path', required=True)
parser.add_argument('--glide_path', required=True)
args = parser.parse_args()
for iter in args.iter.split(','):
iter=int(iter)
iter_gen(args,iter)