rdkit.Chem.MCS

Source Code for Module rdkit.Chem.MCS

1 # This work was funded by Roche and generously donated to the free 2 # and open source cheminformatics community. 3 import warnings 4 warnings.simplefilter('default', DeprecationWarning) 5 warnings.warn("The rdkit.Chem.MCS module is deprecated; please use rdkit.Chem.rdFMCS instead.", 6 DeprecationWarning,stacklevel=2) 7 ## Copyright (c) 2012 Andrew Dalke Scientific AB 8 ## Andrew Dalke <dalke@dalkescientific.com> 9 ## 10 ## All rights reserved. 11 ## 12 ## Redistribution and use in source and binary forms, with or without 13 ## modification, are permitted provided that the following conditions are 14 ## met: 15 ## 16 ## * Redistributions of source code must retain the above copyright 17 ## notice, this list of conditions and the following disclaimer. 18 ## 19 ## * Redistributions in binary form must reproduce the above copyright 20 ## notice, this list of conditions and the following disclaimer in 21 ## the documentation and/or other materials provided with the 22 ## distribution. 23 ## 24 ## THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 25 ## "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 26 ## LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 27 ## A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 28 ## HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 29 ## SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 30 ## LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 31 ## DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 32 ## THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 33 ## (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 34 ## OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 35 from rdkit.Chem import fmcs 36 from rdkit.Chem.fmcs import Default 37 """MCS - find a Maximum Common Substructure 38 39 This software finds the maximum common substructure of a set of 40 structures and reports it as a SMARTS string. 41 42 The SMARTS string depends on the desired match properties. For 43 example, if ring atoms are only allowed to match ring atoms then an 44 aliphatic ring carbon in the query is converted to the SMARTS "[C;R]", 45 and the double-bond ring bond converted to "=;@" while the respective 46 chain-only version are "[C;!R]" and "=;!@". 47 48 """ 49 50 # The simplified algorithm description is: 51 # 52 # best_substructure = None 53 # pick one structure as the query, and other as the targets 54 # for each substructure in the query graph: 55 # convert it to a SMARTS string based on the desired match properties 56 # if the SMARTS pattern exists in all of the targets: 57 # then this is a common substructure 58 # keep track of the maximum such common structure, 59 # 60 # The algorithm will usually take a long time. There are several 61 # ways to speed it up. 62 # 63 # == Bond elimination == 64 # 65 # As the first step, remove bonds which obviously cannot be part of the 66 # MCS. 67 # 68 # This requires atom and bond type information, which I store as SMARTS 69 # patterns. A bond can only be in the MCS if its canonical bond type is 70 # present in all of the structures. A bond type is string made of the 71 # SMARTS for one atom, the SMARTS for the bond, and the SMARTS for the 72 # other atom. The canonical bond type is the lexographically smaller of 73 # the two possible bond types for a bond. 74 # 75 # The atom and bond SMARTS depend on the type comparison used. 76 # 77 # The "ring-matches-ring-only" option adds an "@" or "!@" to the bond 78 # SMARTS, so that the canonical bondtype for "C-C" becomes [#6]-@[#6] or 79 # [#6]-!@[#6] if the bond is in a ring or not in a ring, and if atoms 80 # are compared by element and bonds are compared by bondtype. (This 81 # option does not add "R" or "!R" to the atom SMARTS because there 82 # should be a single bond in the MCS of c1ccccc1O and CO.) 83 # 84 # The result of all of this atom and bond typing is a "TypedMolecule" 85 # for each input structure. 86 # 87 # I then find which canonical bondtypes are present in all of the 88 # structures. I convert each TypedMolecule into a 89 # FragmentedTypedMolecule which has the same atom information but only 90 # those bonds whose bondtypes are in all of the structures. This can 91 # break a structure into multiple, disconnected fragments, hence the 92 # name. 93 # 94 # (BTW, I would like to use the fragmented molecules as the targets 95 # because I think the SMARTS match would go faster, but the RDKit SMARTS 96 # matcher doesn't like them. I think it's because the new molecule 97 # hasn't been sanitized and the underlying data structure the ring 98 # information doesn't exist. Instead, I use the input structures for the 99 # SMARTS match.) 100 # 101 # == Use the structure with the smallest largest fragment as the query == 102 # == and sort the targets by the smallest largest fragment == 103 # 104 # I pick one of the FragmentedTypedMolecule instances as the source of 105 # substructure enumeration. Which one? 106 # 107 # My heuristic is to use the one with the smallest largest fragment. 108 # Hopefully it produces the least number of subgraphs, but that's also 109 # related to the number of rings, so a large linear graph will product 110 # fewer subgraphs than a small fused ring system. I don't know how to 111 # quantify that. 112 # 113 # For each of the fragmented structures, I find the number of atoms in 114 # the fragment with the most atoms, and I find the number of bonds in 115 # the fragment with the most bonds. These might not be the same 116 # fragment. 117 # 118 # I sort the input structures by the number of bonds in the largest 119 # fragment, with ties broken first on the number of atoms, and then on 120 # the input order. The smallest such structure is the query structure, 121 # and the remaining are the targets. 122 # 123 # == Use a breadth-first search and a priority queue to == 124 # == enumerate the fragment subgraphs == 125 # 126 # I extract each of the fragments from the FragmentedTypedMolecule into 127 # a TypedFragment, which I use to make an EnumerationMolecule. An 128 # enumeration molecule contains a pair of directed edges for each atom, 129 # which simplifies the enumeration algorithm. 130 # 131 # The enumeration algorithm is based around growing a seed. A seed 132 # contains the current subgraph atoms and bonds as well as an exclusion 133 # set of bonds which cannot be used for future grown. The initial seed 134 # is the first bond in the fragment, which may potentially grow to use 135 # the entire fragment. The second seed is the second bond in the 136 # fragment, which is excluded from using the first bond in future 137 # growth. The third seed starts from the third bond, which may not use 138 # the first or second bonds during growth, and so on. 139 # 140 # 141 # A seed can grow along bonds connected to an atom in the seed but which 142 # aren't already in the seed and aren't in the set of excluded bonds for 143 # the seed. If there are no such bonds then subgraph enumeration ends 144 # for this fragment. Given N bonds there are 2**N-1 possible ways to 145 # grow, which is just the powerset of the available bonds, excluding the 146 # no-growth case. 147 # 148 # This breadth-first growth takes into account all possibilties of using 149 # the available N bonds so all of those bonds are added to the exclusion 150 # set of the newly expanded subgraphs. 151 # 152 # For performance reasons, the bonds used for growth are separated into 153 # 'internal' bonds, which connect two atoms already in the subgraph, and 154 # 'external' bonds, which lead outwards to an atom not already in the 155 # subgraph. 156 # 157 # Each seed growth can add from 0 to N new atoms and bonds. The goal is 158 # to maximize the subgraph size so the seeds are stored in a priority 159 # queue, ranked so the seed with the most bonds is processed first. This 160 # turns the enumeration into something more like a depth-first search. 161 # 162 # 163 # == Prune seeds which aren't found in all of the structures == 164 # 165 # At each stage of seed growth I check that the new seed exists in all 166 # of the original structures. (Well, all except the one which I 167 # enumerate over in the first place; by definition that one will match.) 168 # If it doesn't match then there's no reason to include this seed or any 169 # larger seeds made from it. 170 # 171 # The check is easy; I turn the subgraph into its corresponding SMARTS 172 # string and use RDKit's normal SMARTS matcher to test for a match. 173 # 174 # There are three ways to generate a SMARTS string: 1) arbitrary, 2) 175 # canonical, 3) hybrid. 176 # 177 # I have not tested #1. During most of the development I assumed that 178 # SMARTS matches across a few hundred structures would be slow, so that 179 # the best solution is to generate a *canonical* SMARTS and cache the 180 # match information. 181 # 182 # Well, it turns out that my canonical SMARTS match code takes up most 183 # of the MCS run-time. If I drop the canonicalization step then the 184 # code averages about 5-10% faster. This isn't the same as #1 - I still 185 # do the initial atom assignment based on its neighborhood, which is 186 # like a circular fingerprint of size 2 and *usually* gives a consistent 187 # SMARTS pattern, which I can then cache. 188 # 189 # However, there are times when the non-canonical SMARTS code is slower. 190 # Obviously one is if there are a lot of structures, and another if is 191 # there is a lot of symmetry. I'm still working on characterizing this. 192 # 193 # 194 # == Maximize atoms? or bonds? == 195 # 196 # The above algorithm enumerates all subgraphs of the query and 197 # identifies those subgraphs which are common to all input structures. 198 # 199 # It's trivial then to keep track of the current "best" subgraph, which 200 # can defined as having the subgraph with the most atoms, or the most 201 # bonds. Both of those options are implemented. 202 # 203 # It would not be hard to keep track of all other subgraphs which are 204 # the same size. 205 # 206 # == complete_ring_only implementation == 207 # 208 # The "complete ring only" option is implemented by first enabling the 209 # "ring-matches-ring-only" option, as otherwise it doesn't make sense. 210 # 211 # Second, in order to be a "best" subgraph, all bonds in the subgraph 212 # which are ring bonds in the original molecule must also be in a ring 213 # in the subgraph. This is handled as a post-processing step. 214 # 215 # (Note: some possible optimizations, like removing ring bonds from 216 # structure fragments which are not in a ring, are not yet implemented.) 217 # 218 # 219 # == Prune seeds which have no potential for growing large enough == 220 # 221 # Given a seed, its set of edges available for growth, and the set of 222 # excluded bonds, figure out the maximum possible growth for the seed. 223 # If this maximum possible is less than the current best subgraph then 224 # prune. 225 # 226 # This requires a graph search, currently done in Python, which is a bit 227 # expensive. To speed things up, I precompute some edge information. 228 # That is, if I know that a given bond is a chain bond (not in a ring) 229 # then I can calculate the maximum number of atoms and bonds for seed 230 # growth along that bond, in either direction. However, precomputation 231 # doesn't take into account the excluded bonds, so after a while the 232 # predicted value is too high. 233 # 234 # Again, I'm still working on characterizing this, and an implementation 235 # in C++ would have different tradeoffs. 236 237 __all__ = ["FindMCS"] 238 239 ########## Main driver for the MCS code 240 241

242 -class MCSResult(object):

243

244 - def __init__(self, obj):

245 self.numAtoms = obj.num_atoms 246 self.numBonds = obj.num_bonds 247 self.smarts = obj.smarts 248 self.completed = obj.completed

249

250 - def __nonzero__(self):

251 return self.smarts is not None

252

253 - def __repr__(self):

254 return "MCSResult(numAtoms=%d, numBonds=%d, smarts=%r, completed=%d)" % ( 255 self.numAtoms, self.numBonds, self.smarts, self.completed)

256

257 - def __str__(self):

258 msg = "MCS %r has %d atoms and %d bonds" % (self.smarts, self.numAtoms, self.numBonds) 259 if not self.completed: 260 msg += " (timed out)" 261 return msg

262 263

264 -def FindMCS(mols, 265 minNumAtoms=2, 266 maximize=Default.maximize, 267 atomCompare=Default.atomCompare, 268 bondCompare=Default.bondCompare, 269 matchValences=Default.matchValences, 270 ringMatchesRingOnly=False, 271 completeRingsOnly=False, 272 timeout=Default.timeout, 273 threshold=None, ):

274 """Find the maximum common substructure of a set of molecules 275 276 In the simplest case, pass in a list of molecules and get back 277 an MCSResult object which describes the MCS: 278 279 >>> from rdkit import Chem 280 >>> mols = [Chem.MolFromSmiles("C#CCP"), Chem.MolFromSmiles("C=CCO")] 281 >>> from rdkit.Chem import MCS 282 >>> MCS.FindMCS(mols) 283 MCSResult(numAtoms=2, numBonds=1, smarts='[#6]-[#6]', completed=1) 284 285 The SMARTS '[#6]-[#6]' matches the largest common substructure of 286 the input structures. It has 2 atoms and 1 bond. If there is no 287 MCS which is at least `minNumAtoms` in size then the result will set 288 numAtoms and numBonds to -1 and set smarts to None. 289 290 By default, two atoms match if they are the same element and two 291 bonds match if they have the same bond type. Specify `atomCompare` 292 and `bondCompare` to use different comparison functions, as in: 293 294 >>> MCS.FindMCS(mols, atomCompare="any") 295 MCSResult(numAtoms=3, numBonds=2, smarts='[*]-[*]-[*]', completed=1) 296 >>> MCS.FindMCS(mols, bondCompare="any") 297 MCSResult(numAtoms=3, numBonds=2, smarts='[#6]~[#6]~[#6]', completed=1) 298 299 An atomCompare of "any" says that any atom matches any other atom, 300 "elements" compares by element type, and "isotopes" matches based on 301 the isotope label. Isotope labels can be used to implement user-defined 302 atom types. A bondCompare of "any" says that any bond matches any 303 other bond, and "bondtypes" says bonds are equivalent if and only if 304 they have the same bond type. 305 306 A substructure has both atoms and bonds. The default `maximize` 307 setting of "atoms" finds a common substructure with the most number 308 of atoms. Use maximize="bonds" to maximize the number of bonds. 309 Maximizing the number of bonds tends to maximize the number of rings, 310 although two small rings may have fewer bonds than one large ring. 311 312 You might not want a 3-valent nitrogen to match one which is 5-valent. 313 The default `matchValences` value of False ignores valence information. 314 When True, the atomCompare setting is modified to also require that 315 the two atoms have the same valency. 316 317 >>> MCS.FindMCS(mols, matchValences=True) 318 MCSResult(numAtoms=2, numBonds=1, smarts='[#6v4]-[#6v4]', completed=1) 319 320 It can be strange to see a linear carbon chain match a carbon ring, 321 which is what the `ringMatchesRingOnly` default of False does. If 322 you set it to True then ring bonds will only match ring bonds. 323 324 >>> mols = [Chem.MolFromSmiles("C1CCC1CCC"), Chem.MolFromSmiles("C1CCCCCC1")] 325 >>> MCS.FindMCS(mols) 326 MCSResult(numAtoms=7, numBonds=6, smarts='[#6]-[#6]-[#6]-[#6]-[#6]-[#6]-[#6]', completed=1) 327 >>> MCS.FindMCS(mols, ringMatchesRingOnly=True) 328 MCSResult(numAtoms=4, numBonds=3, smarts='[#6](-@[#6])-@[#6]-@[#6]', completed=1) 329 330 You can further restrict things and require that partial rings 331 (as in this case) are not allowed. That is, if an atom is part of 332 the MCS and the atom is in a ring of the entire molecule then 333 that atom is also in a ring of the MCS. Set `completeRingsOnly` 334 to True to toggle this requirement and also sets ringMatchesRingOnly 335 to True. 336 337 >>> mols = [Chem.MolFromSmiles("CCC1CC2C1CN2"), Chem.MolFromSmiles("C1CC2C1CC2")] 338 >>> MCS.FindMCS(mols) 339 MCSResult(numAtoms=6, numBonds=6, smarts='[#6]-1-[#6]-[#6](-[#6])-[#6]-1-[#6]', completed=1) 340 >>> MCS.FindMCS(mols, ringMatchesRingOnly=True) 341 MCSResult(numAtoms=5, numBonds=5, smarts='[#6]-@1-@[#6]-@[#6](-@[#6])-@[#6]-@1', completed=1) 342 >>> MCS.FindMCS(mols, completeRingsOnly=True) 343 MCSResult(numAtoms=4, numBonds=4, smarts='[#6]-@1-@[#6]-@[#6]-@[#6]-@1', completed=1) 344 345 The MCS algorithm will exhaustively search for a maximum common substructure. 346 Typically this takes a fraction of a second, but for some comparisons this 347 can take minutes or longer. Use the `timeout` parameter to stop the search 348 after the given number of seconds (wall-clock seconds, not CPU seconds) and 349 return the best match found in that time. If timeout is reached then the 350 `completed` property of the MCSResult will be 0 instead of 1. 351 352 >>> mols = [Chem.MolFromSmiles("Nc1ccccc1"*100), Chem.MolFromSmiles("Nc1ccccccccc1"*100)] 353 >>> MCS.FindMCS(mols, timeout=0.1) 354 MCSResult(..., completed=0) 355 356 (The MCS after 50 seconds contained 511 atoms.) 357 """ 358 warnings.warn("The rdkit.Chem.MCS module is deprecated; please use rdkit.Chem.rdFMCS instead.", 359 DeprecationWarning,stacklevel=2) 360 361 ores = fmcs.fmcs(mols, 362 minNumAtoms=minNumAtoms, 363 maximize=maximize, 364 atomCompare=atomCompare, 365 bondCompare=bondCompare, 366 threshold=threshold, 367 matchValences=matchValences, 368 ringMatchesRingOnly=ringMatchesRingOnly, 369 completeRingsOnly=completeRingsOnly, 370 timeout=timeout, ) 371 return MCSResult(ores)

372 373 374 #------------------------------------ 375 # 376 # doctest boilerplate 377 #

378 -def _test():

379 import doctest, sys 380 return doctest.testmod(sys.modules["__main__"], 381 optionflags=doctest.ELLIPSIS + doctest.NORMALIZE_WHITESPACE)

382 383 384 if __name__ == '__main__': 385 import sys 386 failed, tried = _test() 387 sys.exit(failed) 388