generated from pypa/sampleproject
-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
326 lines (250 loc) · 11.2 KB
/
app.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
""" app.py """
import os
import time
from typing import List
from deepchem.models.normalizing_flows import NormalizingFlowModel
import joblib
import numpy as np
import pandas as pd
import selfies as sf
import streamlit as st
import tensorflow as tf
from rdkit import Chem
from rdkit.Chem import Descriptors as descriptors
from rdkit.Chem import Draw
from tensorflow import keras
from main import get_solubility_parameters, sample
from main import get_normalizing_flow_layer, get_selfies_alphabet
from src.utils.evaluations import tanimoto_similarity
main_container = st.container()
generative_container = st.container()
@st.cache(allow_output_mutation=True)
def load_sklearn_models(filename: str):
path_to_model = os.path.join(os.getcwd(), filename)
with open(file=path_to_model, mode="rb") as file:
model = joblib.load(filename=file)
return model
@st.cache(allow_output_mutation=True)
def load_tensorflow_models(filename: str):
path_to_model = os.path.join(os.getcwd(), filename)
model = tf.saved_model.load(path_to_model)
return model
@st.cache(allow_output_mutation=True)
def load_normalizing_flow_model():
nf = get_normalizing_flow_layer(dim=2000)
nfm = NormalizingFlowModel(nf, learning_rate = 1e-4, batch_size = 128, model_dir="model/generative/flow/generative-normalizing-flow")
nfm.restore()
return nfm
def get_image(molecule):
return Draw.MolToImage(molecule)
def reset():
change_wgan_state()
st.session_state["molecules"] = []
st.session_state["img_index"] = 0
if "img_index" not in st.session_state:
st.session_state["img_index"] = 0
if "gvae_index" not in st.session_state:
st.session_state["gvae_index"] = 0
if "wgan_generated" not in st.session_state:
st.session_state["wgan_generated"] = []
if "gvae_generated" not in st.session_state:
st.session_state["gvae_generated"] = False
if "molecules" not in st.session_state:
st.session_state["molecules"] = []
if "gvae_molecules" not in st.session_state:
st.session_state["gvae_molecules"] = []
if "qm9" not in st.session_state:
data = pd.read_csv("data/qm9.csv")["smiles"].sample(5000, random_state=42)
st.session_state["QM9"] = data.apply(Chem.MolFromSmiles)
def change_wgan_state():
st.session_state["wgan_generated"] = True
def change_gvae_state():
st.session_state["gvae_generated"] = True
solubility_scaler = load_sklearn_models(
"model/regressors/solubility-standard-scaler-sklearn.model"
)
solubility_model = load_sklearn_models("model/regressors/solubility-rf-sklearn.model")
with main_container:
st.title("Generative modelling of small molecules")
st.write(
"In this project we aim to use generative models to design small molecules and predict their solubility"
)
st.write(
"""
1. Generative models
- Graph Variational Autoencoder
- Wasserstein GAN
- Normalizing Generative Flow
2. Solubility prediction
- Random Forest
"""
)
with generative_container:
model = st.radio(
"Choose a generative model.", ("", "GAN", "VAE", "NF"), horizontal=True, index=0
)
# TODO: Fix bug in
progress_bar = generative_container.progress(0)
st.text(
"Tanimoto Similarity is calculated using 5000 molecules from the QM9 database"
)
if model == "GAN":
wgan = load_tensorflow_models("model/generative/gans/qm9/2023-01-05_12-58-04")
if st.button("Reset WGAN", key="wgan_reset"):
st.session_state["wgan_generated"] = False
st.session_state["index"] = 0
if not st.session_state["wgan_generated"]:
molecules = sample(wgan.generator, model_type="WGAN")
st.session_state["molecules"] = [
molecule for molecule in molecules if molecule
]
change_wgan_state()
for perc_completed in range(100):
time.sleep(0.001)
progress_bar.progress(perc_completed + 1)
molecules = (
st.session_state["molecules"] if st.session_state["molecules"] else []
)
if molecules:
smiles = [Chem.MolToSmiles(mol.GetMol()) for mol in molecules if mol]
index = st.session_state["img_index"]
st.text("WGAN generated molecules using QM9 dataset")
prevb, _, nextb = st.columns([1, 5, 1])
if prevb.button("Previous", key="prev"):
if index - 1 < 0:
print(st.session_state["img_index"])
st.session_state["img_index"] = len(smiles)
else:
st.session_state["img_index"] = st.session_state["img_index"] - 1
if nextb.button("Next", key="next"):
if st.session_state["img_index"] > len(smiles) - 1:
st.session_state["img_index"] = 0
else:
st.session_state["img_index"] = index + 1
img_col, props_col = st.columns(2)
img = get_image(molecules[index])
img_col.text(f"Molecule {index+1} of {len(smiles)} ")
img_col.image(img, caption=smiles[index])
descriptor = Chem.MolFromSmiles(smiles[index])
predictors = np.array([get_solubility_parameters(descriptor)])
predicted_solubility = solubility_model.predict(
solubility_scaler.transform(predictors)
)[0]
data = st.session_state["QM9"]
props_col.text("Physico-Chemical Properties")
props_col.text(f"SMILES Descriptor: {smiles[index]}")
props_col.text(
f"Molecular Weight: {descriptors.MolWt(descriptor):.3f} g/mol"
)
props_col.text(f"log(Solubility): {predicted_solubility:.3f} mol/L")
mean_tanimoto: float = np.mean(tanimoto_similarity(data, descriptor))
std_tanimoto: float = np.std(tanimoto_similarity(data, descriptor))
props_col.text(
f"Mean Tanimoto Similarity: {mean_tanimoto:.3f} +/- {std_tanimoto:.3f}"
)
else:
st.text("No valid molecule was generated. Try Again!")
elif model == "VAE":
gvae = load_tensorflow_models("model/generative/vaes/zinc/2023-01-05_16-53-59")
if st.button("Reset GVAE", key="gvae_reset"):
st.session_state["gvae_generated"] = False
st.session_state["gvae_index"] = 0
if not st.session_state["gvae_generated"]:
molecules = sample(model=gvae.decoder, model_type="GVAE", batch_size=50)
st.session_state["gvae_molecules"] = [
molecule for molecule in molecules if molecule
]
change_gvae_state()
for perc_completed in range(100):
time.sleep(0.001)
progress_bar.progress(perc_completed + 1)
molecules = (
st.session_state["gvae_molecules"]
if st.session_state["gvae_molecules"]
else []
)
if molecules:
index = st.session_state["gvae_index"]
smiles = [Chem.MolToSmiles(mol.GetMol()) for mol in molecules if mol]
st.text("Generating small molecules using GVAE...")
prevb, _, nextb = st.columns([1, 5, 1])
if prevb.button("Previous", key="prev"):
if index - 1 < 0:
st.session_state["gvae_index"] = len(smiles)
else:
st.session_state["gvae_index"] = st.session_state["gvae_index"] - 1
if nextb.button("Next", key="next"):
if index > len(smiles) - 1:
st.session_state["gvae_index"] = 0
else:
st.session_state["gvae_index"] = index + 1
img_col, props_col = st.columns(2)
img = get_image(molecules[index])
img_col.text(f"Generated {index+1} out of {len(molecules)} molecules")
img_col.image(img, caption=smiles[index])
descriptor = Chem.MolFromSmiles(smiles[0])
predictors = np.array([get_solubility_parameters(descriptor)])
predicted_solubility = solubility_model.predict(
solubility_scaler.transform(predictors)
)[0]
data = st.session_state["QM9"]
props_col.text("Physico-Chemical Properties")
props_col.text(f"SMILES Descriptor: {smiles[index]}")
props_col.text(
f"Molecular Weight: {descriptors.MolWt(descriptor):.3f} g/mol"
)
props_col.text(f"log(Solubility): {predicted_solubility:.3f} mol/L")
mean_tanimoto: float = np.mean(tanimoto_similarity(data, descriptor))
std_tanimoto: float = np.std(tanimoto_similarity(data, descriptor))
props_col.text(
f"Mean Tanimoto Similarity: {mean_tanimoto:.3f} +/- {std_tanimoto:.3f}"
)
else:
st.text("No valid molecule was generated. Try Again!")
elif model == "NF":
nfm = load_normalizing_flow_model()
generated_samples = nfm.flow.sample(20)
log_probs = nfm.flow.log_probs(generated_samples)
mols = tf.math.floor(generated_samples)
mols = tf.clip_by_value(mols, 0, 1)
int_to_symbol = dict((i, c) for i, c in enumerate(get_selfies_alphabet()))
mols = mols.numpy().tolist()
selfies_molecule = sf.encoding_to_selfies(mols, vocab_itos=int_to_symbol, enc_type="one_hot")
smiles = sf.decoder(selfies_molecule)
molecules = [Chem.SmilesFromMol(mol) for mol in smiles]
st.text("Generating small molecules using Normalizing Flow...")
index = 0
prevb, _, nextb = st.columns([1, 5, 1])
if prevb.button("Previous", key="prev"):
if index - 1 < 0:
st.session_state["gvae_index"] = len(smiles)
else:
st.session_state["gvae_index"] = st.session_state["gvae_index"] - 1
if nextb.button("Next", key="next"):
if index > len(smiles) - 1:
st.session_state["gvae_index"] = 0
else:
st.session_state["gvae_index"] = index + 1
img_col, props_col = st.columns(2)
img = get_image(molecules[index])
img_col.text(f"Generated {index+1} out of {len(molecules)} molecules")
img_col.image(img, caption=smiles[index])
descriptor = Chem.MolFromSmiles(smiles[0])
predictors = np.array([get_solubility_parameters(descriptor)])
predicted_solubility = solubility_model.predict(
solubility_scaler.transform(predictors)
)[0]
data = st.session_state["QM9"]
props_col.text("Physico-Chemical Properties")
props_col.text(f"SMILES Descriptor: {smiles[index]}")
props_col.text(
f"Molecular Weight: {descriptors.MolWt(descriptor):.3f} g/mol"
)
props_col.text(f"log(Solubility): {predicted_solubility:.3f} mol/L")
mean_tanimoto: float = np.mean(tanimoto_similarity(data, descriptor))
std_tanimoto: float = np.std(tanimoto_similarity(data, descriptor))
props_col.text(
f"Tanimoto Similarity: {mean_tanimoto:.3f} +/- {std_tanimoto:.3f}"
)
else:
pass