File: AtomAtomPathSimilarity.py

package info (click to toggle)
rdkit 201809.1%2Bdfsg-6
  • links: PTS, VCS
  • area: main
  • in suites: buster
  • size: 123,688 kB
  • sloc: cpp: 230,509; python: 70,501; java: 6,329; ansic: 5,427; sql: 1,899; yacc: 1,739; lex: 1,243; makefile: 445; xml: 229; fortran: 183; sh: 123; cs: 93
file content (380 lines) | stat: -rw-r--r-- 14,098 bytes parent folder | download
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
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
from __future__ import print_function
import numpy
import time
import unittest

from scipy.optimize import linear_sum_assignment

from rdkit import Chem
from rdkit.Chem import AllChem
from rdkit.Chem import rdmolops
from rdkit import DataStructs
from rdkit.Chem.Fingerprints import FingerprintMols

_BK_ = {
  Chem.rdchem.BondType.SINGLE: 1,
  Chem.rdchem.BondType.DOUBLE: 2,
  Chem.rdchem.BondType.TRIPLE: 3,
  Chem.rdchem.BondType.AROMATIC: 4
}
_BONDSYMBOL_ = {1: '-', 2: '=', 3: '#', 4: ':'}

#_nAT_ = 217 # 108*2+1
_nAT_ = 223  # Gobbi code actually uses the first prime higher than 217, not 217 itself
_nBT_ = 5

#def FindAllPathsOfLengthN_Gobbi(mol, length, rootedAtAtom=-1, uniquepaths=True):
#	return FindAllPathsOfLengthMToN(mol, length, length, rootedAtAtom=rootedAtAtom, uniquepaths=uniquepaths)


def FindAllPathsOfLengthMToN_Gobbi(mol, minlength, maxlength, rootedAtAtom=-1, uniquepaths=True):
  '''this function returns the same set of bond paths as the Gobbi paper.  These differ a little from the rdkit FindAllPathsOfLengthMToN function'''
  paths = []
  for atom in mol.GetAtoms():
    if rootedAtAtom == -1 or atom.GetIdx() == rootedAtAtom:
      path = []
      visited = set([atom.GetIdx()])
      #			visited = set()
      _FindAllPathsOfLengthMToN_Gobbi(atom, path, minlength, maxlength, visited, paths)

  if uniquepaths:
    uniquepathlist = []
    seen = set()
    for path in paths:
      if path not in seen:
        reversepath = tuple([i for i in path[::-1]])
        if reversepath not in seen:
          uniquepathlist.append(path)
          seen.add(path)
    return uniquepathlist
  else:
    return paths


def _FindAllPathsOfLengthMToN_Gobbi(atom, path, minlength, maxlength, visited, paths):
  for bond in atom.GetBonds():
    if bond.GetIdx() not in path:
      bidx = bond.GetIdx()
      path.append(bidx)
      if len(path) >= minlength and len(path) <= maxlength:
        paths.append(tuple(path))
      if len(path) < maxlength:
        a1 = bond.GetBeginAtom()
        a2 = bond.GetEndAtom()
        if a1.GetIdx() == atom.GetIdx():
          nextatom = a2
        else:
          nextatom = a1
        nextatomidx = nextatom.GetIdx()
        if nextatomidx not in visited:
          visited.add(nextatomidx)
          _FindAllPathsOfLengthMToN_Gobbi(nextatom, path, minlength, maxlength, visited, paths)
          visited.remove(nextatomidx)
      path.pop()


def getpathintegers(m1, uptolength=7):
  '''returns a list of integers describing the paths for molecule m1.  This uses numpy 16 bit unsigned integers to reproduce the data in the Gobbi paper.  The returned list is sorted'''
  bondtypelookup = {}
  for b in m1.GetBonds():
    bondtypelookup[b.GetIdx()] = _BK_[b.GetBondType()], b.GetBeginAtom(), b.GetEndAtom()
  pathintegers = {}
  for a in m1.GetAtoms():
    idx = a.GetIdx()
    pathintegers[idx] = []
    #		for pathlength in range(1, uptolength+1):
    #			for path in rdmolops.FindAllPathsOfLengthN(m1, pathlength, rootedAtAtom=idx):
    for ipath, path in enumerate(
        FindAllPathsOfLengthMToN_Gobbi(m1, 1, uptolength, rootedAtAtom=idx, uniquepaths=False)):
      strpath = []
      currentidx = idx
      res = []
      for ip, p in enumerate(path):
        bk, a1, a2 = bondtypelookup[p]
        strpath.append(_BONDSYMBOL_[bk])
        if a1.GetIdx() == currentidx:
          a = a2
        else:
          a = a1
        ak = a.GetAtomicNum()
        if a.GetIsAromatic():
          ak += 108
#trying to get the same behaviour as the Gobbi test code - it looks like a circular path includes the bond, but not the closure atom - this fix works
        if a.GetIdx() == idx:
          ak = None
        if ak is not None:
          astr = a.GetSymbol()
          if a.GetIsAromatic():
            strpath.append(astr.lower())
          else:
            strpath.append(astr)
        res.append((bk, ak))
        currentidx = a.GetIdx()
      pathuniqueint = numpy.ushort(0)  # work with 16 bit unsigned integers and ignore overflow...
      for ires, (bi, ai) in enumerate(res):
        #use 16 bit unsigned integer arithmetic to reproduce the Gobbi ints
        #					pathuniqueint = ((pathuniqueint+bi)*_nAT_+ai)*_nBT_
        val1 = pathuniqueint + numpy.ushort(bi)
        val2 = val1 * numpy.ushort(_nAT_)
        #trying to get the same behaviour as the Gobbi test code - it looks like a circular path includes the bond, but not the closure atom - this fix works
        if ai is not None:
          val3 = val2 + numpy.ushort(ai)
          val4 = val3 * numpy.ushort(_nBT_)
        else:
          val4 = val2
        pathuniqueint = val4
      pathintegers[idx].append(pathuniqueint)
#sorted lists allow for a quicker comparison algorithm
  for p in pathintegers.values():
    p.sort()
  return pathintegers


def getcommon(l1, ll1, l2, ll2):
  '''returns the number of items sorted lists l1 and l2 have in common.  ll1 and ll2 are the list lengths'''
  ncommon = 0
  ix1 = 0
  ix2 = 0
  while (ix1 < ll1) and (ix2 < ll2):
    a1 = l1[ix1]
    a2 = l2[ix2]
    #a1 is < or > more often that ==
    if a1 < a2:
      ix1 += 1
    elif a1 > a2:
      ix2 += 1
    else:  # a1 == a2:
      ncommon += 1
      ix1 += 1
      ix2 += 1
  return ncommon


def getsimaibj(aipaths, bjpaths, naipaths, nbjpaths):
  '''returns the similarity of two sorted path lists.  Equation 2'''
  nc = getcommon(aipaths, naipaths, bjpaths, nbjpaths)
  sim = float(nc + 1) / (max(naipaths, nbjpaths) * 2 - nc + 1)
  return sim


def getmappings(simmatrixarray):
  '''return a mapping of the atoms in the similarity matix using the heuristic algorithm described in the paper'''

  costarray = numpy.ones(simmatrixarray.shape) - simmatrixarray

  it = numpy.nditer(costarray, flags=['multi_index'], op_flags=['writeonly'])
  dsu = []
  for a in it:
    dsu.append((a, it.multi_index[0], it.multi_index[1]))
  dsu.sort()

  seena = set()
  seenb = set()
  mappings = []
  for sim, a, b in dsu:
    if a not in seena and b not in seenb:
      seena.add(a)
      seenb.add(b)
      mappings.append((a, b))

  return mappings[:min(simmatrixarray.shape)]


def gethungarianmappings(simmatrixarray):
  '''return a mapping of the atoms in the similarity matrix - the Hungarian algorithm is used because it is invariant to atom ordering.  Requires scipy'''
  costarray = numpy.ones(simmatrixarray.shape) - simmatrixarray
  row_ind, col_ind = linear_sum_assignment(costarray)
  res = zip(row_ind, col_ind)
  return res


def getsimab(mappings, simmatrixdict):
  '''return the similarity for a set of mapping.  See Eqn 3'''
  naa, nab = simmatrixdict.shape

  score = 0.0
  for a, b in mappings:
    score += simmatrixdict[a][b]
  simab = score / (max(naa, nab) * 2 - score)
  return simab


def getsimmatrix(m1, m1pathintegers, m2, m2pathintegers):
  '''generate a matrix of atom atom similarities.  See Figure 4'''

  aidata = [((ai.GetAtomicNum(), ai.GetIsAromatic()), ai.GetIdx()) for ai in m1.GetAtoms()]
  bjdata = [((bj.GetAtomicNum(), bj.GetIsAromatic()), bj.GetIdx()) for bj in m2.GetAtoms()]

  simmatrixarray = numpy.zeros((len(aidata), len(bjdata)))

  for ai, (aitype, aiidx) in enumerate(aidata):
    aipaths = m1pathintegers[aiidx]
    naipaths = len(aipaths)
    for bj, (bjtype, bjidx) in enumerate(bjdata):
      if aitype == bjtype:
        bjpaths = m2pathintegers[bjidx]
        nbjpaths = len(bjpaths)
        simmatrixarray[ai][bj] = getsimaibj(aipaths, bjpaths, naipaths, nbjpaths)
  return simmatrixarray


def AtomAtomPathSimilarity(m1, m2, m1pathintegers=None, m2pathintegers=None):
  '''compute the Atom Atom Path Similarity for a pair of RDKit molecules.  See Gobbi et al, J. ChemInf (2015) 7:11
	the most expensive part of the calculation is computing the path integers - we can precompute these and pass them in as an argument'''
  if m1pathintegers is None:
    m1pathintegers = getpathintegers(m1)
  if m2pathintegers is None:
    m2pathintegers = getpathintegers(m2)

  simmatrix = getsimmatrix(m1, m1pathintegers, m2, m2pathintegers)

  #	mappings = getmappings(simmatrix)
  mappings = gethungarianmappings(simmatrix)

  simab = getsimab(mappings, simmatrix)

  return simab


def test0():
  '''reproduce the worked similarity in the Gobbi paper'''
  m1 = Chem.MolFromSmiles('o1nccc1C')
  m2 = Chem.MolFromSmiles('[nH]1nccc1')
  return AtomAtomPathSimilarity(m1, m2)


def test1():
  '''generate a set of path integers for 3 molecules from the Gobbi source IAAPathGeneratorCharTest.java'''
  res = []
  smiles = ["C", "C(=O)F", "C1ON1"]
  for s in smiles:
    m = Chem.MolFromSmiles(s)
    mpathintegers = getpathintegers(m)
    res.append(mpathintegers)
  return res


def test2():
  '''generate a matrix molecules from the Gobbi source AAPathComparator2Test.java'''
  smileslist = ["*", "C", "N", "CCO", "CC(=O)N", "c1ccccc1", "c1ncncc1", "c1[nH]ccc1",
                "c1ncncc1CC(=O)N", "c1ccccc1c1ncncc1"]
  sims = []
  for s1 in smileslist:
    for s2 in smileslist:
      m1 = Chem.MolFromSmiles(s1)
      m2 = Chem.MolFromSmiles(s2)
      sims.append('%.4f' % AtomAtomPathSimilarity(m1, m2))
  return sims


def test3():
  '''generate a set of similarities for the example compounds in Figure 1.  These are compared to the values in Additional File 1'''
  m1a = Chem.MolFromSmiles('Clc1ccc(CN2CCC(CC2)c3cc([nH]n3)c4ccc(Cl)cc4)cc1')
  m1b = Chem.MolFromSmiles('Clc1ccc(CN2CCN(CC2)CC(=O)N(C)c3ccccc3)cc1')
  m2a = Chem.MolFromSmiles('Cc1cccn2cc(nc12)c3ccc(NC(=O)CN4CCCC4)cc3')
  m2b = Chem.MolFromSmiles('Cc1c(cc2ccccn12)c3ccc(OCCCN4CCCCC4)cc3')

  res = []
  for m1 in (m1a, m1b, m2a, m2b):
    for m2 in (m1a, m1b, m2a, m2b):
      sim = AtomAtomPathSimilarity(m1, m2)
      res.append('%.2f' % sim)
  return res


def timeit():
  #these are the first 40 smiles in the Gobbi 13321_2015_56_MOESM2_ESM file'''
  molstr = '''C[C@@H](O)[C@@H]1OCC[C@@H](C)[C@H](O)C(=O)OC[C@]23CCC(C)=C[C@H]2O[C@@H]4C[C@@H](OC(=O)C=CC=C1)[C@@]3(C)[C@@]45CO5
CC1=C[C@H]2O[C@@H]3C[C@H]4OC(=O)C=CC=CC(=O)OCCC(C)=CC(=O)OC[C@@]2(CC1)[C@]4(C)[C@]35CO5
CC1=CC(=O)OC[C@]23C[C@H](O)C(C)=C[C@H]2O[C@@H]4C[C@@H](OC(=O)C=CC=CC(=O)OCC1)[C@@]3(C)[C@]45CO5
CC1(C)N=C(N)N=C(N)N1C2=CC=C(Br)C=C2
CC1(C)N=C(N)N=C(N)N1C2=CC=CC=C2
CC1(C)N=C(N)N=C(N)N1C2=CC=C(I)C=C2
CC1=CC=C(C=C1)N2C(N)=NC(N)=NC2(C)C
CC1(C)N=C(N)N=C(N)N1C2=CC=C(Cl)C=C2
CC1(C)N=C(N)N=C(N)N1C2=CC=C(F)C=C2
CC1=CC=CC(=C1)N2C(N)=NC(N)=NC2(C)C
COC1=CC=C(C=C1)N2C(N)=NC(N)=NC2(C)C
CC1=CC=C(N2C(N)=NC(N)=NC2(C)C)C(C)=C1
CCOC1=CC=C(C=C1)N2C(N)=NC(N)=NC2(C)C
COC1=CC=CC(=C1)N2C(N)=NC(N)=NC2(C)C
CC1=CC=CC(NC2=NC(N)=NC(C)(C)N2)=C1
CNC1=C(N(CC2=CC=C(Cl)C(Cl)=C2)C(C)=O)C(=O)C3=CC=CC=C3C1=O
CC(=O)N(CC1=CC=C(F)C=C1)C2=C(NCC3=CC=CC=C3)C(=O)C4=CC=CC=C4C2=O
CC(=O)N(CC1=CC=C(F)C=C1)C2=C(NCCC3=CC=CC=C3)C(=O)C4=CC=CC=C4C2=O
CCC(=O)N(C(C)C)C1=C(NC)C(=O)C2=CC=CC=C2C1=O
CCN(CC)CCCC(C)NC1=CC=NC2=CC(Cl)=CC=C12
CC(CCCNCCO)NC1=CC=NC2=CC(Cl)=CC=C12
CC(C)C(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
CC(C)(C)CC(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
CC(C)CC(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
CC(CC(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12)CC(C)(C)C
CCCCC(CC)C(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
ClC1=CC=C2C(NCCCCNC(=O)C3CCCC3)=CC=NC2=C1
CCCCCCCCC(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
CC(C)(CCl)C(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
ClC1=CC=C2C(NCCCCNC(=O)C3CCCCC3)=CC=NC2=C1
NCCCCCCNC1=CC=NC2=CC(Cl)=CC=C12
ClC1=CC=C2C(NCCCCNC(=O)CCC3CCCC3)=CC=NC2=C1
ClC1=CC=C2C(NC3CCCCCCC3)=CC=NC2=C1
CC1CCC(CC1)NC2=CC=NC3=CC(Cl)=CC=C23
CN(C)C(=O)NCCCCNC1=CC=NC2=CC(Cl)=CC=C12
CCN(CC)CCCC(C)NC1=C2C=CC(Cl)=CC2=NC3=CC=C(OC)C=C13
CC(CCCO)NC1=CC=NC2=CC(Cl)=CC=C12
CCCCCC(=O)NCCCCCCNC1=CC=NC2=CC(Cl)=CC=C12
ClC1=CC=C2C(NC3CCC(CC3)NC(=O)CCC4CCCC4)=CC=NC2=C1
CN(C)CCCNC1=CC=NC2=CC(Cl)=CC=C12
'''

  mols = [Chem.MolFromSmiles(smiles) for smiles in molstr.splitlines()]
  na = len(mols)
  nb = len(mols)
  pathints = [getpathintegers(mol) for mol in mols]
  start = time.time()
  for a, api in zip(mols, pathints):
    for b, bpi in zip(mols, pathints):
      sim = AtomAtomPathSimilarity(a, b, m1pathintegers=api, m2pathintegers=bpi)
  print('time to compute %dx%d matrix: %.2fs' % (na, nb, time.time() - start))


class TestAtomAtomPathSimilarity(unittest.TestCase):

  def test_paper(self):
    self.assertEqual('%.3f' % test0(), '0.066')

  def test_getcommon(self):
    self.assertEqual(getcommon([2, 2, 2, 3, 3, 3], 6, [1, 2, 3, 3, 4, 5], 6), 3)

  def test_pathintegers(self):
    self.assertEqual(test1(), [
      {0: []}, {0: [1160, 2270],
                1: [2260, 30692],
                2: [1145, 33761]}, {0: [752, 1150, 1155, 3826, 38221, 43791],
                                    1: [1145, 1150, 1596, 4670, 32641, 38211],
                                    2: [1145, 1155, 5785, 32646, 43786, 65173]}
    ])

  def test_AAPathComparator2Test(self):
    self.assertEqual(
      test2(),
      ['1.0000', '0.0000', '0.0000', '0.0000', '0.0000', '0.0000', '0.0000', '0.0000', '0.0000',
       '0.0000', '0.0000', '1.0000', '0.0000', '0.0345', '0.0182', '0.0000', '0.0000', '0.0000',
       '0.0017', '0.0000', '0.0000', '0.0000', '1.0000', '0.0000', '0.0182', '0.0000', '0.0000',
       '0.0000', '0.0020', '0.0000', '0.0000', '0.0345', '0.0000', '1.0000', '0.1126', '0.0000',
       '0.0000', '0.0000', '0.0088', '0.0000', '0.0000', '0.0182', '0.0182', '0.1126', '1.0000',
       '0.0000', '0.0000', '0.0000', '0.0336', '0.0000', '0.0000', '0.0000', '0.0000', '0.0000',
       '0.0000', '1.0000', '0.0373', '0.0645', '0.0148', '0.0869', '0.0000', '0.0000', '0.0000',
       '0.0000', '0.0000', '0.0373', '1.0000', '0.1101', '0.1767', '0.0869', '0.0000', '0.0000',
       '0.0000', '0.0000', '0.0000', '0.0645', '0.1101', '1.0000', '0.0387', '0.0219', '0.0000',
       '0.0017', '0.0020', '0.0088', '0.0336', '0.0148', '0.1767', '0.0387', '1.0000', '0.0869',
       '0.0000', '0.0000', '0.0000', '0.0000', '0.0000', '0.0869', '0.0869', '0.0219', '0.0869',
       '1.0000'])

  def test_tableS1(self):
    self.assertEqual(test3(), ['1.00', '0.19', '0.06', '0.09', '0.19', '1.00', '0.12', '0.05',
                               '0.06', '0.12', '1.00', '0.15', '0.09', '0.05', '0.15', '1.00'])


if __name__ == "__main__":
  #	unittest.main()
  timeit()