-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess_data.py
212 lines (174 loc) · 8.78 KB
/
preprocess_data.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
import argparse
import os
import pandas as pd
import random
from rdkit import Chem
from tqdm import tqdm
from tqdm.contrib import tzip
from e_smiles import get_e_smiles, merge_smiles, get_edit_from_e_smiles, merge_smiles_only, get_e_smiles_with_check
from tqdm.contrib.concurrent import process_map
def remove_amap_not_in_product(rxn_smi):
"""
Corrects the atom map numbers of atoms only in reactants.
This correction helps avoid the issue of duplicate atom mapping
after the canonicalization step.
"""
r, p = rxn_smi.split(">>")
pmol = Chem.MolFromSmiles(p)
pmol_amaps = set([atom.GetAtomMapNum() for atom in pmol.GetAtoms()])
max_amap = max(pmol_amaps) #Atoms only in reactants are labelled starting with max_amap
rmol = Chem.MolFromSmiles(r)
for atom in rmol.GetAtoms():
amap_num = atom.GetAtomMapNum()
if amap_num not in pmol_amaps:
atom.SetAtomMapNum(max_amap+1)
max_amap += 1
r_updated = Chem.MolToSmiles(rmol)
rxn_smi_updated = r_updated + ">>" + p
return rxn_smi_updated
def random_smiles_with_map(smiles_with_map):
mol = Chem.MolFromSmiles(smiles_with_map)
atom_map_lis = [a.GetAtomMapNum() for a in mol.GetAtoms()]
[a.ClearProp('molAtomMapNumber') for a in mol.GetAtoms()]
random_id = random.randint(0,len(atom_map_lis)-1)
Chem.Kekulize(mol)
Chem.MolToSmiles(mol,rootedAtAtom = random_id)
order = eval(mol.GetProp("_smilesAtomOutputOrder"))
mol_ordered = Chem.RenumberAtoms(mol, order )
for i in range(len(order)) :
atom = mol_ordered.GetAtomWithIdx(i)
atom.SetAtomMapNum(atom_map_lis[order[i]])
smiles = Chem.MolToSmiles(mol_ordered,canonical = False,kekuleSmiles=True)
return smiles
def canonical_smiles_with_map(smiles_with_map): #unnecessary
mol = Chem.MolFromSmiles(smiles_with_map)
atom_map_lis = [a.GetAtomMapNum() for a in mol.GetAtoms()]
[a.ClearProp('molAtomMapNumber') for a in mol.GetAtoms()]
Chem.Kekulize(mol)
order = list(Chem.CanonicalRankAtoms(mol, includeChirality=True))
mol_ordered = Chem.RenumberAtoms(mol, order )
for i in range(len(order)) :
atom = mol_ordered.GetAtomWithIdx(i)
atom.SetAtomMapNum(atom_map_lis[order[i]])
smiles = Chem.MolToSmiles(mol_ordered,canonical = False,kekuleSmiles=True)
return smiles
def augmentate_data(df, aug_time):
new_dict = {'id': [], 'class': [], 'reactants>reagents>production': []}
# Preprocessing loop
for time in tqdm(range(aug_time), desc = "Augtime"):
for idx in tqdm(range(len(df)), desc = "AugProcess"):
element = df.loc[idx]
uspto_id, class_id, rxn_smi = element['id'], element['class'], element['reactants>reagents>production']
rxn_smi_new = remove_amap_not_in_product(rxn_smi)
r,p = rxn_smi_new.split('>>')
p = random_smiles_with_map(p)
mol = Chem.MolFromSmiles(p,sanitize = False)
old_map_lis = [a.GetAtomMapNum() for a in mol.GetAtoms()]
for i in range(mol.GetNumAtoms()):
mol.GetAtomWithIdx(i).SetAtomMapNum(i + 1)
p = Chem.MolToSmiles(mol,canonical = False,kekuleSmiles=True)
p_mol = Chem.MolFromSmiles(p,sanitize = False)
new_map_lis = [a.GetAtomMapNum() for a in p_mol.GetAtoms()]
if new_map_lis != [i+1 for i in range(len(new_map_lis))]:
print(p)
break
dic_old_new_map = dict(zip(old_map_lis,new_map_lis))
mol = Chem.MolFromSmiles(r)
for i in range(mol.GetNumAtoms()):
atom = mol.GetAtomWithIdx(i)
if atom.GetAtomMapNum() in dic_old_new_map.keys():
atom.SetAtomMapNum(dic_old_new_map[atom.GetAtomMapNum()])
else:
pass
r = Chem.MolToSmiles(mol)
r = canonical_smiles_with_map(r) #unnecessary
rxn_smi_new = r + '>>' + p
new_dict['id'].append(uspto_id)
new_dict['class'].append(class_id)
new_dict['reactants>reagents>production'].append(rxn_smi_new)
new_df = pd.DataFrame.from_dict(new_dict)
return new_df
def preprocess_data(df, data_source, split, augtime):
"""
Clean few data, and Get E-SMILES,
"""
if data_source == '50k':
if split == "train":
rxn_class_list = []
e_smiles_list = []
for class_, rxn in tzip(df['class'], df['reactants>reagents>production']):
try:
o_reactant = rxn.split('>>')[0]
r_mol = Chem.MolFromSmiles(o_reactant)
for atom in r_mol.GetAtoms():
atom.SetAtomMapNum(0)
o_reactant = Chem.MolToSmiles(r_mol)
o_reactant = Chem.MolToSmiles(Chem.MolFromSmiles(o_reactant))
e_smiles = get_e_smiles_with_check(rxn)
n_reactant = merge_smiles_only(e_smiles)
n_reactant = Chem.MolToSmiles(Chem.MolFromSmiles(n_reactant))
if o_reactant == n_reactant:
rxn_class_list.append(f'class_{class_}')
e_smiles_list.append(e_smiles)
else:
# print(o_reactant)
# print(n_reactant)
pass
except:
pass
elif split == "test":
idx_to_drop = [822, 1282, 1490, 1558, 2810, 3487, 4958]
rows_to_drop = []
for j in range(augtime):
rows_to_drop += [j*5007 + i for i in idx_to_drop]
df = df.drop(rows_to_drop)
df = df.reset_index(drop = True)
rxn_class_list = [f"class_{n}" for n in df['class']]
e_smiles_list = process_map(get_e_smiles, tqdm(df['reactants>reagents>production'], desc = "transforming into E-SMILES ..."), max_workers = 20)
elif split == "val":
idx_to_drop = [2302, 2527, 2950, 4368, 4863, 4890]
rows_to_drop = []
for j in range(augtime):
rows_to_drop += [j*5001 + i for i in idx_to_drop]
df = df.drop(rows_to_drop)
df = df.reset_index(drop = True)
rxn_class_list = [f"class_{n}" for n in df['class']]
e_smiles_list = process_map(get_e_smiles, tqdm(df['reactants>reagents>production'], desc = "transforming into E-SMILES ..."), max_workers = 20)
return e_smiles_list, rxn_class_list
def main():
parser = argparse.ArgumentParser(description='Preprocess UPSTO_50k data')
parser.add_argument('-data', type=str, required=True, choices=['50k', 'mit'], help='Data file to preprocess')
parser.add_argument('-split', type=str, required=True, choices=['train', 'val', 'test'], help='Data split to preprocess')
parser.add_argument('-augtime', type=int, required=True, help='Number of augmentations, we set 100 for train and 20 for test')
parser.add_argument('-rxn_class', type=bool, default=False, choices=[False, True], help='Unkown reaction type (False) or Given reaction type (True)')
args = parser.parse_args()
# Load data
raw_df = pd.read_csv(f'./datasets/{args.data}_raw/raw_{args.split}.csv')
# Data Augmentation (Kekulized and Mapped rxn_smiles)
augmentated_df = augmentate_data(raw_df, args.augtime)
# Get E-SMILES, Clean few data
e_smiles_list, rxn_class_list = preprocess_data(augmentated_df, args.data, args.split, args.augtime)
# Tokenization E-SMILES with single characters
src_list, tgt_list = [i.split(">>>")[0] for i in e_smiles_list], [i.split(">>>")[1] for i in e_smiles_list]
# Unknown/Given rxn_class
if not args.rxn_class:
src = [" ".join(list(s)) for s in src_list]
tgt = [" ".join(list(t)) for t in tgt_list]
src_file_path = f"./datasets/{args.data}_ReactSeq/aug_{args.augtime}_{args.split}/src_{args.split}.txt"
tgt_file_path = f"./datasets/{args.data}_ReactSeq/aug_{args.augtime}_{args.split}/tgt_{args.split}.txt"
elif args.rxn_class:
src = [c + " " + " ".join(list(s)) for c,s in zip(rxn_class_list, src_list)]
tgt = [" ".join(list(t)) for t in tgt_list]
src_file_path = f"./datasets/{args.data}_ReactSeq_with_rxn_class/aug_{args.augtime}_{args.split}/src_{args.split}.txt"
tgt_file_path = f"./datasets/{args.data}_ReactSeq_with_rxn_class/aug_{args.augtime}_{args.split}/tgt_{args.split}.txt"
# Write processed data into .txt
os.makedirs(os.path.dirname(src_file_path), exist_ok=True)
os.makedirs(os.path.dirname(tgt_file_path), exist_ok=True)
with open(src_file_path, "w") as f:
for line in src:
f.write(line+'\n')
with open(tgt_file_path, "w") as f:
for line in tgt:
f.write(line+'\n')
if __name__ == '__main__':
main()