1
|
|
|
#! /usr/bin/env python |
2
|
|
|
# |
3
|
|
|
# Copyright (C) 2016 Rich Lewis <[email protected]> |
4
|
|
|
# License: 3-clause BSD |
5
|
|
|
|
6
|
|
|
""" |
7
|
|
|
# skchem.features.descriptors.autocorrelation |
8
|
|
|
|
9
|
|
|
Autocorrelation descriptors for scikit-chem. |
10
|
|
|
""" |
11
|
|
|
|
12
|
|
|
from functools import partial |
13
|
|
|
|
14
|
|
|
from .caching import cache |
|
|
|
|
15
|
|
|
from .fundamentals import adjacency_matrix, distance_matrix, atom_props |
|
|
|
|
16
|
|
|
|
17
|
|
|
import numpy as np |
|
|
|
|
18
|
|
|
|
19
|
|
|
|
20
|
|
|
@cache.inject(distance_matrix, atom_props) |
21
|
|
|
def moreau_broto_autocorrelation(mol, dist_mat, prop, prop_name='atomic_mass', |
|
|
|
|
22
|
|
|
c_scaled=False, centred=False, |
|
|
|
|
23
|
|
|
ks=range(1, 9)): |
24
|
|
|
|
25
|
|
|
""" The Moreau-Broto autocorrelation. |
|
|
|
|
26
|
|
|
|
27
|
|
|
$$ ATS_k = \frac{1}{2} \hdot \sum_{i=1}^A \sum_{j=1}^A $$ |
28
|
|
|
|
29
|
|
|
With special case $$ ATS_0 = \sum_{i=1}^A w_i^2. |
30
|
|
|
|
31
|
|
|
Where $A$ is the number of atoms, and $w$ is an atomic property. |
32
|
|
|
|
33
|
|
|
Args: |
34
|
|
|
mol (skchem.Mol): |
35
|
|
|
The molecule for which to calculate the descriptor. |
36
|
|
|
|
37
|
|
|
prop (str): |
38
|
|
|
The atomic property. |
39
|
|
|
|
40
|
|
|
c_scaled (bool): |
41
|
|
|
Whether the properties should be scaled against sp3 carbon. |
42
|
|
|
|
43
|
|
|
centred (bool): |
44
|
|
|
Whether the descriptor should be divided by the number of |
45
|
|
|
contributions (avoids dependence on molecular size). |
46
|
|
|
|
47
|
|
|
ks (iterable): |
48
|
|
|
The lags to calculate the descriptor over. |
49
|
|
|
|
50
|
|
|
Returns: |
51
|
|
|
float |
52
|
|
|
|
53
|
|
|
Examples: |
54
|
|
|
>>> import skchem |
55
|
|
|
>>> m = skchem.Mol.from_smiles('CC(O)CCO') |
56
|
|
|
>>> moreau_broto_autocorrelation(m, centred=False, |
57
|
|
|
... prop_name='atomic_mass', |
58
|
|
|
... ks=range(4)) # doctest: +ELLIPSIS |
59
|
|
|
array([ 1088.99..., 817.12..., 865.02..., 528.59...]) |
60
|
|
|
""" |
61
|
|
|
|
62
|
|
|
div = np.array([(dist_mat == k).sum() for k in ks]) if centred else 1 |
|
|
|
|
63
|
|
|
|
64
|
|
|
return np.array([(1 if k == 0 else 0.5) * prop.dot(dist_mat == k).dot(prop) |
65
|
|
|
for k in ks]) / div |
66
|
|
|
|
67
|
|
|
|
68
|
|
|
@cache.inject(distance_matrix, atom_props) |
69
|
|
|
def moran_coefficient(mol, dist_mat, prop, prop_name='atomic_mass', |
|
|
|
|
70
|
|
|
c_scaled=False, ks=range(1, 9)): |
|
|
|
|
71
|
|
|
|
72
|
|
|
""" Moran coefficient for lags ks. |
73
|
|
|
|
74
|
|
|
Args: |
75
|
|
|
mol (skchem.Mol): |
76
|
|
|
The molecule for which to calculate the descriptor. |
77
|
|
|
|
78
|
|
|
prop (str): |
79
|
|
|
The atomic property. |
80
|
|
|
|
81
|
|
|
c_scaled (bool): |
82
|
|
|
Whether the properties should be scaled against sp3 carbon. |
83
|
|
|
|
84
|
|
|
centered (bool): |
85
|
|
|
Whether the descriptor should be divided by the number of |
86
|
|
|
contributions (avoids dependence on molecular size. |
87
|
|
|
|
88
|
|
|
ks (iterable): |
89
|
|
|
The lags to calculate the descriptor over. |
90
|
|
|
|
91
|
|
|
Returns: |
92
|
|
|
float |
93
|
|
|
""" |
94
|
|
|
|
95
|
|
|
prop = prop - prop.mean() |
96
|
|
|
|
97
|
|
|
res = [] |
98
|
|
|
|
99
|
|
|
for k in ks: |
|
|
|
|
100
|
|
|
geodesic = dist_mat == k |
101
|
|
|
num = prop.dot(geodesic).dot(prop) / geodesic.sum() |
102
|
|
|
denom = (prop ** 2).sum() / len(prop) |
103
|
|
|
res.append(num / denom) |
104
|
|
|
|
105
|
|
|
return np.array(res) |
106
|
|
|
|
107
|
|
|
|
108
|
|
|
@cache.inject(distance_matrix, atom_props) |
109
|
|
|
def geary_coefficient(mol, dist_mat, prop, prop_name='atomic_mass', |
|
|
|
|
110
|
|
|
c_scaled=False, ks=range(1, 9)): |
|
|
|
|
111
|
|
|
|
112
|
|
|
""" The geary coefficient for *ks* lags. |
113
|
|
|
|
114
|
|
|
Args: |
115
|
|
|
mol (skchem.Mol): |
116
|
|
|
The molecule for which to calculate the descriptor. |
117
|
|
|
|
118
|
|
|
prop (str): |
119
|
|
|
The atomic property. |
120
|
|
|
|
121
|
|
|
c_scaled (bool): |
122
|
|
|
Whether the properties should be scaled against sp3 carbon. |
123
|
|
|
|
124
|
|
|
centered (bool): |
125
|
|
|
Whether the descriptor should be divided by the number of |
126
|
|
|
contributions (avoids dependence on molecular size. |
127
|
|
|
|
128
|
|
|
ks (iterable): |
129
|
|
|
The lags to calculate the descriptor over. |
130
|
|
|
|
131
|
|
|
Returns: |
132
|
|
|
float |
133
|
|
|
|
134
|
|
|
""" |
135
|
|
|
|
136
|
|
|
res = [] |
137
|
|
|
for k in ks: |
|
|
|
|
138
|
|
|
geodesic = dist_mat == k |
139
|
|
|
num = 0.5 * ((prop - prop[:, np.newaxis]) ** 2 * geodesic).sum() / geodesic.sum() |
140
|
|
|
denom = ((prop - prop.mean()) ** 2).sum() / (len(prop) - 1) |
141
|
|
|
res.append(num / denom) |
142
|
|
|
|
143
|
|
|
return np.array(res) |
144
|
|
|
|
145
|
|
|
|
146
|
|
|
@cache |
147
|
|
|
@cache.inject(adjacency_matrix, distance_matrix) |
148
|
|
|
def galvez_matrix(mol, dist_mat, adj_mat): |
149
|
|
|
|
150
|
|
|
""" The galvez matrix. |
151
|
|
|
|
152
|
|
|
Args: |
153
|
|
|
mol (skchem.Mol): |
154
|
|
|
The molecule for which to calculate the matrix. |
155
|
|
|
|
156
|
|
|
Returns: |
157
|
|
|
np.array |
158
|
|
|
""" |
159
|
|
|
|
160
|
|
|
temp = dist_mat ** -2 |
161
|
|
|
np.fill_diagonal(temp, 0) |
162
|
|
|
galvez_mat = temp.dot(adj_mat) + np.diag(mol.atoms.valence_vertex_degree) |
163
|
|
|
return galvez_mat |
164
|
|
|
|
165
|
|
|
|
166
|
|
|
@cache |
167
|
|
|
@cache.inject(galvez_matrix) |
168
|
|
|
def charge_matrix(mol, galvez_mat): |
169
|
|
|
|
170
|
|
|
""" The charge matrix. |
171
|
|
|
|
172
|
|
|
Args: |
173
|
|
|
mol (skchem.Mol): |
174
|
|
|
The molecule for which to calculate the matrix. |
175
|
|
|
|
176
|
|
|
Returns: |
177
|
|
|
np.array |
178
|
|
|
|
179
|
|
|
""" |
180
|
|
|
|
181
|
|
|
ct_mat = galvez_mat - galvez_mat.T |
182
|
|
|
ct_mat[np.diag_indices_from(ct_mat)] = mol.atoms.depleted_degree |
183
|
|
|
|
184
|
|
|
return ct_mat |
185
|
|
|
|
186
|
|
|
|
187
|
|
|
@cache |
188
|
|
|
@cache.inject(charge_matrix, distance_matrix) |
189
|
|
|
def topological_charge_index(mol, c_mat, dist_mat, ks=range(11)): |
|
|
|
|
190
|
|
|
|
191
|
|
|
""" The Galvez tologogical charge index for lags ks. |
192
|
|
|
|
193
|
|
|
Args: |
194
|
|
|
mol (skchem.Mol): |
195
|
|
|
The molecule for which to calculate the descriptor. |
196
|
|
|
|
197
|
|
|
ks (iterable): |
198
|
|
|
The lags for which to calculate the descriptor. |
199
|
|
|
|
200
|
|
|
Returns: |
201
|
|
|
np.array |
202
|
|
|
|
203
|
|
|
""" |
204
|
|
|
return np.array([0.5 * np.abs(c_mat)[dist_mat == k].sum() for i in ks]) |
|
|
|
|
205
|
|
|
|
206
|
|
|
|
207
|
|
|
@cache.inject(topological_charge_index) |
208
|
|
|
def mean_topological_charge_index(mol, tci, ks=range(11)): |
|
|
|
|
209
|
|
|
|
210
|
|
|
""" Mean topological charge index for lags ks. |
211
|
|
|
|
212
|
|
|
Args: |
213
|
|
|
mol (skchem.Mol): |
214
|
|
|
The molecule for which to calculate the descriptor. |
215
|
|
|
|
216
|
|
|
ks (iterable): |
217
|
|
|
The lags for which to calculate the descriptor. |
218
|
|
|
|
219
|
|
|
Returns: |
220
|
|
|
np.array |
221
|
|
|
""" |
222
|
|
|
|
223
|
|
|
return tci / (len(mol.atoms) - 1) |
224
|
|
|
|
225
|
|
|
|
226
|
|
|
@cache.inject(topological_charge_index) |
227
|
|
|
def total_charge_index(mol, tci, ks=range(11)): |
|
|
|
|
228
|
|
|
|
229
|
|
|
return tci.sum() / (len(mol.atoms) - 1) |
230
|
|
|
|
231
|
|
|
|
232
|
|
|
PROPS = ['atomic_mass', 'van_der_waals_volume', 'sanderson_electronegativity', |
233
|
|
|
'polarisability', 'ionisation_energy', 'intrinsic_state'] |
234
|
|
|
|
235
|
|
|
KS = range(1, 9) |
236
|
|
|
|
237
|
|
|
DESCRIPTORS = [partial(moreau_broto_autocorrelation, k=k, prop=p, centered=c) |
238
|
|
|
for k in KS for c in (False, True) for p in PROPS] |
239
|
|
|
|
240
|
|
|
FS = (moran_coefficient, geary_coefficient) |
241
|
|
|
|
242
|
|
|
DESCRIPTORS += [partial(f, k=k, prop=p) for f in FS for k in KS for p in PROPS] |
243
|
|
|
|
244
|
|
|
__all__ = ['moreau_broto_autocorrelation', 'moran_coefficient', |
245
|
|
|
'geary_coefficient', 'topological_charge_index', |
246
|
|
|
'mean_topological_charge_index', 'total_charge_index'] |
247
|
|
|
|
This can be caused by one of the following:
1. Missing Dependencies
This error could indicate a configuration issue of Pylint. Make sure that your libraries are available by adding the necessary commands.
2. Missing __init__.py files
This error could also result from missing
__init__.py
files in your module folders. Make sure that you place one file in each sub-folder.