This repository has been archived by the owner on Jun 22, 2024. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
203 lines (189 loc) · 8.72 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
# -*- coding: utf-8 -*-
# Author: Nianze A. Tao (Omomzawa Sueno)
"""
Web UI
"""
import math
import argparse
from rdkit.Chem import Draw, MolFromSmiles
from mol2chemfigPy3 import mol2chemfig
import gradio as gr
import torch
from torch.jit import load, freeze
from akane2.utils.graph import smiles2graph, gather
from akane2.utils.token import protein2vec
chkpt = "torchscript_model"
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model1 = freeze(load(f"{chkpt}/moleculenet/esol.pt").eval().to(device))
model2 = freeze(load(f"{chkpt}/moleculenet/freesolv.pt").eval().to(device))
model3 = freeze(load(f"{chkpt}/moleculenet/lipo.pt").eval().to(device))
model4 = freeze(load(f"{chkpt}/qm9/qm9_homo.pt").eval().to(device))
model5 = freeze(load(f"{chkpt}/qm9/qm9_lumo.pt").eval().to(device))
model6 = freeze(load(f"{chkpt}/qm9/qm9_gap.pt").eval().to(device))
model7 = freeze(load(f"{chkpt}/qm9/qm9_zpve.pt").eval().to(device))
model8 = freeze(load(f"{chkpt}/qm9/qm9_u0.pt").eval().to(device))
model9 = freeze(load(f"{chkpt}/qm9/qm9_u.pt").eval().to(device))
model10 = freeze(load(f"{chkpt}/qm9/qm9_g.pt").eval().to(device))
model11 = freeze(load(f"{chkpt}/qm9/qm9_h.pt").eval().to(device))
model12 = freeze(load(f"{chkpt}/qm9/qm9_cv.pt").eval().to(device))
model13 = freeze(load(f"{chkpt}/photoswitch/e_iso_n.pt").eval().to(device))
model14 = freeze(load(f"{chkpt}/photoswitch/e_iso_pi.pt").eval().to(device))
model15 = freeze(load(f"{chkpt}/photoswitch/z_iso_n.pt").eval().to(device))
model16 = freeze(load(f"{chkpt}/photoswitch/z_iso_pi.pt").eval().to(device))
model17 = freeze(load(f"{chkpt}/protein_ligand.pt").eval().to(device))
model18 = freeze(load(f"{chkpt}/deep_eutectic_solvent.pt").eval().to(device))
def _count_iso(mol):
counter = []
bonds = mol.GetBonds()
for bond in bonds:
type = int(bond.GetBondType())
if type == 2:
stereo = int(bond.GetStereo())
if stereo == 2 or stereo == 4:
counter.append("Z")
elif stereo == 3 or stereo == 5:
counter.append("E")
else:
atom_b, atom_e = bond.GetBeginAtom(), bond.GetEndAtom()
degree_b, degree_e = atom_b.GetDegree(), atom_e.GetDegree()
if degree_b != 1 and degree_e != 1:
counter.append("any")
return set(counter)
@torch.no_grad()
def process0(smiles: str):
mol = MolFromSmiles(smiles)
img = Draw.MolToImage(mol, (500, 500))
counter = _count_iso(mol)
mol = gather([smiles2graph(smiles)])
mol["node"] = mol["node"].to(device)
mol["edge"] = mol["edge"].to(device)
v1 = math.pow(10, model1(mol)["prediction"].item())
v2 = model2(mol)["prediction"].item()
v3 = model3(mol)["prediction"].item()
v4 = model4(mol)["prediction"].item()
v5 = model5(mol)["prediction"].item()
v6 = model6(mol)["prediction"].item()
v7 = model7(mol)["prediction"].item()
v8 = model8(mol)["prediction"].item()
v9 = model9(mol)["prediction"].item()
v10 = model10(mol)["prediction"].item()
v11 = model11(mol)["prediction"].item()
v12 = model12(mol)["prediction"].item()
if "E" in counter:
v13 = model13(mol)["prediction"].item()
v14 = model14(mol)["prediction"].item()
elif "Z" in counter:
v13 = model15(mol)["prediction"].item()
v14 = model16(mol)["prediction"].item()
elif "any" in counter:
v13 = v14 = "please specify the double bond stereochemistry in SMILES."
else:
v13 = v14 = "n.a."
return img, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14
@torch.no_grad()
def process1(size, file):
with open(file.name, "r", encoding="utf-8") as f:
data = f.readlines()
fasta = data[1]
label = torch.tensor([protein2vec(fasta)], device=device)
while True:
smiles = model17.generate(int(size), label)
mol = MolFromSmiles(smiles["SMILES"])
if mol != None:
break
img = Draw.MolToImage(mol, (500, 500))
smiles = smiles["SMILES"]
chemfig = mol2chemfig(smiles, "-r", inline=True)
return img, smiles, chemfig
@torch.no_grad()
def process2(size, x, mp):
label = torch.tensor([[float(x), float(mp)]], device=device)
while True:
smiles = model18.generate(int(size), label)
mol = MolFromSmiles(smiles["SMILES"])
if mol != None:
break
img = Draw.MolToImage(mol, (500, 500))
smiles = smiles["SMILES"]
chemfig = mol2chemfig(smiles, "-r", inline=True)
return img, smiles, chemfig
with gr.Blocks(title="AkAne") as app:
gr.Markdown(
"### This model is part of MSc Electrochemistry and Battery Technologies project (2022 - 2023), University of Southampton"
)
gr.Markdown("Author: Nianze Tao (Omozawa Sueno)")
gr.Markdown("---")
with gr.Tab(label="MOLECULAR PROPERTY PREDICTION"):
with gr.Row():
with gr.Column(scale=1):
smiles0 = gr.Textbox(label="SMILES", placeholder="input SMILES here")
btn0 = gr.Button("RUN", variant="primary")
img0 = gr.Image(label="molecule")
with gr.Column(scale=2):
with gr.Tab(label="EXP"):
gr.Markdown(
"Predicted with A<span style='color:#CB4154'>k</span>Ane from MoleculeNet training data."
)
v1 = gr.Textbox(label="solubility / M")
v2 = gr.Textbox(label="solvation energy / kcal/mol")
v3 = gr.Textbox(label="lipophilicity (logD)")
gr.Markdown(
"Predicted with A<span style='color:#CB4154'>k</span>Ane from PhotoSwitch training data."
)
v13 = gr.Textbox(label="n-π* transition wavelength / nm")
v14 = gr.Textbox(label="π-π* transition wavelength / nm")
with gr.Tab(label="QM"):
gr.Markdown(
"Predicted with A<span style='color:#CB4154'>k</span>Ane from QM9 training data."
)
v4 = gr.Textbox(label="HOMO / Hartree")
v5 = gr.Textbox(label="LUMO / Hartree")
v6 = gr.Textbox(label="HOMO-LUMO gap / Hartree")
v7 = gr.Textbox(label="zero-point vibrational energy / Hartree")
v8 = gr.Textbox(label="U at 0K / Hartree")
v9 = gr.Textbox(label="U at 298.15K / Hartree")
v10 = gr.Textbox(label="G at 298.15K / Hartree")
v11 = gr.Textbox(label="H at 298.15K / Hartree")
v12 = gr.Textbox(label="heat capacity / kcal/K/mol")
with gr.Tab(label="DE NOVO STRUCTURE GENERATION"):
with gr.Tab(label="protein ligand design"):
gr.Markdown(
"*de novo* ligand desiged with A<span style='color:#CB4154'>k</span>Ane from BindingDB training data."
)
with gr.Row():
with gr.Column(scale=1):
size1 = gr.Slider(10, 100, value=24, step=1, label="molecule size")
file = gr.File(file_types=[".fasta"], label="FASTA sequence")
btn1 = gr.Button("RUN", variant="primary")
img1 = gr.Image(label="molecule")
with gr.Column(scale=2):
smiles1 = gr.Textbox(label="SMILES")
chemfig1 = gr.TextArea(label="LATEX ChemFig")
with gr.Tab(label="deep eutective solvent design"):
gr.Markdown(
"*de novo* deep eutective solvent desiged with A<span style='color:#CB4154'>k</span>Ane."
)
with gr.Row():
with gr.Column(scale=1):
size2 = gr.Slider(5, 100, value=12, step=1, label="molecule size")
x = gr.Slider(0.1, 0.9, label="x(HBA)", value=0.5)
mp = gr.Slider(
-120, 200, label="melting point / °C", value=-20, step=0.2
)
btn2 = gr.Button("RUN", variant="primary")
img2 = gr.Image(label="molecule")
with gr.Column(scale=2):
smiles2 = gr.Textbox(label="SMILES")
chemfig2 = gr.TextArea(label="LATEX ChemFig")
btn0.click(
fn=process0,
inputs=smiles0,
outputs=[img0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14],
)
btn1.click(fn=process1, inputs=[size1, file], outputs=[img1, smiles1, chemfig1])
btn2.click(fn=process2, inputs=[size2, x, mp], outputs=[img2, smiles2, chemfig2])
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--public", default=False, type=bool, help="open to public")
args = parser.parse_args()
app.launch(share=args.public)