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
|
# -*- coding: utf-8 -*-
"""
Implementation of Ertl's Functional Group Algorithm in RDKit
Ertl, Peter.
An algorithm to identify functional groups in organic molecules
J Cheminform (2017) 9:36; DOI: 10.1186/s13321-017-0225-z
@author: gonzalo.colmenarejo
"""
from rdkit import Chem
from rdkit.Chem.Draw import IPythonConsole
from chembl_structure_pipeline import standardizer as sdz
from datetime import datetime
from IPython.display import Image
from PIL import Image as pilImage
from PIL import ImageDraw, ImageFont
from rdkit.Chem.Draw import rdMolDraw2D
import efgs
import pandas as pd
import io
ertl_smis = [
'Cc1nc(NS(=O)(=O)c2ccc(N)cc2)nc(C)c1',
'NC(=N)c1ccc(C=Cc2ccc(cc2O)C(=N)N)cc1',
'CC(=O)Nc1nnc(s1)S(=O)(=O)N',
'NS(=O)(=O)c1cc2c(NCNS2(=O)=O)cc1Cl',
'CNC1=Nc2ccc(Cl)cc2C(=N(=O)C1)c3ccccc3',
'Cc1onc(c1C(=O)NC2C3SC(C)(C)C(N3C2=O)C(=O)O)c4ccccc4',
'Clc1ccccc1C2=NCC(=O)Nc3ccc(cc23)N(=O)=O',
'COc1cc(cc(C(=O)NCC2CCCN2CC=C)c1OC)S(=O)(=O)N',
'Cc1ccc(Cl)c(Nc2ccccc2C(=O)O)c1Cl',
'Clc1ccc2Oc3ccccc3N=C(N4CCNCC4)c2c1',
'FC(F)(F)CN1C(=O)CN=C(c2ccccc2)c3cc(Cl)ccc13',
'OCC1OC(CC1O)n2cnc3C(O)CNC=Nc32',
'CCNC1CC(C)S(=O)(=O)c2sc(cc12)S(=O)(=O)N',
'CC(O)C1C2C(C)C(=C(N2C1=O)C(=O)O)SC3CNC(C3)C(=O)N(C)C',
'CC1CN(CC(C)N1)c2c(F)c(N)c3c(=O)c(cn(C4CC4)c3c2F)C(=O)O',
'CC(=CCC1C(=O)N(N(C1=O)c2ccccc2)c3ccccc3)C',
'Clc1ccc2N=C3NC(=O)CN3Cc2c1Cl',
'CC(=O)NC1C(NC(=N)N)C=C(OC1C(O)C(O)CO)C(=O)O',
'CC(O)C(O)C1CNc2nc(N)nc(O)c2N1',
'NC1CCCCN(C1)c2c(Cl)cc3c(=O)c(cn(C4CC4)c3c2Cl)C(=O)O',
]
def mindex(mol):
# Utility function to draw molecule with idx
for atom in mol.GetAtoms():
try:
atom.SetAtomMapNum(atom.GetIdx())
except:
None
return mol
def demindex(mol):
# Utility function to return mindex-ed molecule to normal
for atom in mol.GetAtoms():
atom.ClearProp('molAtomMapNumber')
return mol
def molimg(x):
# Utility function to draw image of labeled mol from PNG binary string
try:
return pilImage.open(io.BytesIO(x))
except:
return None
def img_grid(images, titles = None, num_columns = 5,font = "arial.ttf", font_size = 18, title_height = 30, cell_width = 300, cell_height = 300, bg_color = (255, 255, 255)):
"""
Creates a grid of images with optional titles under each image.
Parameters:
- images (list): List of PIL Image objects.
- titles (list, optional): List of titles as strings for each image. If None, no titles will be added.
- num_columns (int): Number of columns in the grid.
- font_sizes: Font size of titles (if provided).
- title_height: Height of titles (if provieded).
- cell_width (int): Width of each image cell.
- cell_height (int): Height of each image cell (not including title space).
- bg_color (tuple): Background color as an RGB tuple.
Returns:
- PIL Image object containing the image grid.
"""
# Set title height if titles are provided; otherwise, set to zero
title_height = title_height if titles else 0
cell_height_with_title = cell_height + title_height + 2
# Calculate grid dimensions
num_images = len(images)
num_rows = (num_images + num_columns - 1) // num_columns # Round up for last row
# Create a blank background image for the grid
grid_width = num_columns * cell_width
grid_height = num_rows * cell_height_with_title
grid_image = pilImage.new("RGB", (grid_width, grid_height), bg_color)
# Set up font for titles if needed
if titles:
try:
font = ImageFont.truetype(font, font_size) # Change font and size as desired
except IOError:
font = ImageFont.load_default()
# Iterate through each cell in the grid
for i in range(num_rows * num_columns):
row = i // num_columns
col = i % num_columns
x = col * cell_width
y = row * cell_height_with_title
if i < num_images:
# Resize the image to fit in the cell
img_resized = images[i].resize((cell_width, cell_height))
grid_image.paste(img_resized, (x, y))
# Draw title if provided
if titles and i < len(titles):
draw = ImageDraw.Draw(grid_image)
title = titles[i]
# Calculate title position and center it
text_bbox = draw.textbbox((0, 0), title, font=font)
text_width = text_bbox[2] - text_bbox[0]
title_x = x + (cell_width - text_width) // 2 # Center title horizontally
title_y = y + cell_height # Position directly beneath the image
# Draw title text
draw.text((title_x, title_y), title, fill="black", font=font)
else:
# Leave blank space if no image for this cell
blank_cell = pilImage.new("RGB", (cell_width, cell_height), bg_color)
grid_image.paste(blank_cell, (x, y))
return grid_image
# ## Mols from file
start_time = datetime.now()
d = pd.read_csv("../data/ch33_inchis.csv", sep = ";", low_memory = False)
d = d[pd.notna(d.inchi)].reset_index(drop = True) # make sure there are no missing inchis
d = d[d.inchi.apply(lambda x: "." not in x.split("/")[1])].reset_index(drop = True) # make sure we have no complex molecules
d.shape # (839063, 1)
d = d.sample(frac=1, random_state=4210).reset_index(drop = True) # scramble the df
n = d.shape[0] # Here one can put an alternative much smaller n
d = d[d.index.isin(range(n))].reset_index(drop = True) # keep only n
d["mol"] = d.inchi.apply(lambda x: sdz.standardize_mol(Chem.MolFromInchi(x)))
d = d[pd.notna(d.mol)].reset_index(drop = True) # remove entries without molecule object
d.shape[0] # n
end_time = datetime.now()
print('Duration to analyze {} molecules: {}'.format(n, end_time - start_time))
start_time = datetime.now()
d[["img_text","fgs","fgsmi","fgmol"]] = d.apply(lambda x: pd.Series(efgs.get_dec_fgs(x["mol"])), axis = 1)
end_time = datetime.now()
print('Duration to analyze {} molecules: {}'.format(n, end_time - start_time))
# Generate distribution of FGs
fg_dist = d.explode(["fgsmi","fgmol"]).groupby("fgsmi", as_index = False).apply(lambda x:
pd.DataFrame({"fgsmi": [x["fgsmi"].iloc[0]],
"n": [x.shape[0]],
"mol": [x["fgmol"].iloc[0]]})).sort_values("n", ascending = False)
fg_dist.shape # 5829, 3 (for all ChEMBL comps)
fg_dist.to_csv("../results/fg_dist.csv", sep = ";", index = False)
# fg_dist = pd.read_csv("../results/fg_dist.csv", sep = ";", index = False) # To recover data
mols = fg_dist.mol.iloc[:64].tolist()
legs = fg_dist.fgsmi.iloc[:64].tolist()
## High-resolution grid with top-48 FGs
images = []
img_width = 250
img_height = 250
for m in mols[:48]:
Chem.rdDepictor.Compute2DCoords(m)
Chem.rdDepictor.StraightenDepiction(m)
d2d = rdMolDraw2D.MolDraw2DCairo(img_width,img_height)
dopts = d2d.drawOptions()
dopts.bondLineWidth = 5
dopts.padding = 0.2
d2d.DrawMolecule(m)
d2d.FinishDrawing()
images.append(molimg(d2d.GetDrawingText()))
fgs_grid = img_grid(images[:48], titles = legs[:48], num_columns = 8, font_size = 28, cell_width = 300, cell_height = 300, bg_color=(255, 255, 255))
fgs_grid = fgs_grid.resize((6000, int(6000*fgs_grid.height/fgs_grid.width)))
fgs_grid.save("../results/fg_grid.png", dpi = (300,300))
# High-resolution grid with 20 first molecules with colored functional groups
images = [molimg(x) for x in d.img_text.iloc[0:20].tolist()]
cell_width = images[0].width
cell_height = images[0].height
molfg_grid = img_grid(images, num_columns = 5, cell_width = cell_width, cell_height = cell_height, bg_color=(255, 255, 255))
molfg_grid
molfg_grid = molfg_grid.resize((6000, int(6000*molfg_grid.height/molfg_grid.width)))
molfg_grid.save("../RESULTS/molfg_grid.png", dpi = (300,300))
# High-resolution grid with 20 Ertl molecules in Figure 1
ertmols = [Chem.MolFromSmiles(s) for s in ertl_smis[:20]]
images = [molimg(efgs.get_dec_fgs(ertmol)[0]) for ertmol in ertmols]
cell_width = images[0].width
cell_height = images[0].height
molfg_grid = img_grid(images, num_columns = 5, cell_width = cell_width, cell_height = cell_height, bg_color=(255, 255, 255))
molfg_grid
molfg_grid = molfg_grid.resize((6000, int(6000*molfg_grid.height/molfg_grid.width)))
molfg_grid.save("../RESULTS/ertl_molfg_grid.png", dpi = (300,300))
|
