def check_mol( mol ):
match, mList = EmbedLib.MatchPharmacophoreToMol( mol, feat_fact, pcophore)
if match:
res = []
num_match = len( mList )
for i in range( num_match ):
num_feature = len( mList[i] )
for j in range( num_feature ):
print mList[i][j].GetAtomIds(), mList[i][j].GetFamily()
bounds = rdDistGeom.GetMoleculeBoundsMatrix( mol )
pList = EmbedLib.GetAllPharmacophoreMatches( mList, bounds, pcophore )
#pList = EmbedLib.MatchPharmacophore( mList, bounds, pcophore )
print pList
#print raw_input("-----")
num_match = len( pList )
print num_match
phMatches = []
for i in range( num_match ):
num_feature = len( pList[i] )
phMatch = []
for j in range( num_feature ):
phMatch.append( pList[i][j].GetAtomIds() )
phMatches.append( phMatch )
for phMatch in phMatches:
bm, embeds, nFail = EmbedLib.EmbedPharmacophore( mol, phMatch, pcophore, count=20, silent=1 )
print "-----> embeds num:", len( embeds )
for embed in embeds:
AllChem.UFFOPtimizeMolecule( embed )
align_data = rdAliginment.GetAlignmetTransform( bm, bounds )
AllChem.TransformMol( embed, align_data[1] )
res.append( embed )
return res
def _align_molecules(self, molecules: List[Chem.Mol]) -> None:
""" Align a list of molecules to a given pharmacophore.
Parameters
----------
molecules : list of rdkit.Chem.Mol
List of molecules to align.
"""
self.n_molecules += len(molecules)
rdkit_pharmacophore, radii = self.pharmacophore.to_rdkit()
apply_radii_to_bounds(radii, rdkit_pharmacophore)
fdef = os.path.join(RDConfig.RDDataDir, 'BaseFeatures.fdef')
featFactory = ChemicalFeatures.BuildFeatureFactory(fdef)
MolScore = namedtuple("MolScore", ["score", "id", "mol"])
for mol in tqdm(molecules):
bounds_matrix = rdDistGeom.GetMoleculeBoundsMatrix(mol)
can_match, all_matches = EmbedLib.MatchPharmacophoreToMol(
mol, featFactory, rdkit_pharmacophore)
if can_match:
failed, _, matched_mols, _ = EmbedLib.MatchPharmacophore(
all_matches,
bounds_matrix,
rdkit_pharmacophore,
useDownsampling=True)
if failed:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
else:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
atom_match = [list(x.GetAtomIds()) for x in matched_mols]
try:
mol_H = Chem.AddHs(mol)
_, embeddings, _ = EmbedLib.EmbedPharmacophore(
mol_H, atom_match, rdkit_pharmacophore, count=10)
except:
continue
SSDs = transform_embeddings(rdkit_pharmacophore, embeddings,
atom_match)
if len(SSDs) == 0:
matched_mol = MolScore(0.0, mol.GetProp("_Name"), mol)
self.molecules.append(matched_mol)
continue
best_fit_index = min(enumerate(SSDs), key=itemgetter(1))[0]
score = 1 / SSDs[best_fit_index]
matched_mol = MolScore(score, mol.GetProp("_Name"),
embeddings[best_fit_index])
self.molecules.append(matched_mol)
def test3Embed(self):
testResults = {
'mol_197':
(218.80, 35.75, 110.33, 11.58, 109.66, 11.09, 90.35, 2.95, 0.00),
'mol_223':
(259.19, 6.27, 134.13, 1.12, 134.06, 1.12, 85.74, 0.61, 0.00),
'mol_269':
(204.51, 7.89, 103.89, 1.20, 102.66, 1.20, 88.07, 1.21, 6.00),
}
inF = gzip.open(os.path.join(self.dataDir, 'cdk2-syn-clip100.pkl.gz'),
'rb')
nDone = 0
nHits = 0
while 1:
try:
name, molPkl, _ = cPickle.load(inF, encoding='latin1')
if PY3:
molPkl = bytes(molPkl, encoding='latin1')
except Exception:
break
nDone += 1
mol = Chem.Mol(molPkl)
nboundsMat = rdDistGeom.GetMoleculeBoundsMatrix(mol)
DG.DoTriangleSmoothing(nboundsMat)
matched, matches = EmbedLib.MatchPharmacophoreToMol(
mol, self.featFactory, self.pcophore)
if matched:
failed, _, match, stats = EmbedLib.MatchPharmacophore(
matches, nboundsMat, self.pcophore, useDownsampling=1)
if not failed:
nHits += 1
if name in testResults:
stats = EmbedLib.EmbedOne(mol,
name,
match,
self.pcophore,
count=10,
silent=1,
randomSeed=23)
tgt = testResults[name]
self.assertEqual(len(tgt), len(stats))
print(name)
print(','.join(['%.2f' % x for x in stats]))
# we'll use different tolerances for the different values:
self.assertTrue(feq(tgt[0], stats[0], 5.0),
(tgt[0], stats[0]))
for i in range(2, len(tgt)):
self.assertTrue(feq(tgt[i], stats[i], 5.0),
(tgt[i], stats[i]))
self.assertEqual(nDone, 100)
# print 'nHits:',nHits
self.assertEqual(nHits, 50)
def test4Search(self):
featFactory = ChemicalFeatures.BuildFeatureFactory(
os.path.join(self.dataDir, 'BaseFeatures.fdef'))
activeFeats = [ChemicalFeatures.FreeChemicalFeature('Acceptor',
Geometry.Point3D(0.0, 0.0, 0.0)),
ChemicalFeatures.FreeChemicalFeature('Donor', Geometry.Point3D(0.0, 0.0, 0.0)),
ChemicalFeatures.FreeChemicalFeature('Aromatic',
Geometry.Point3D(0.0, 0.0, 0.0))]
pcophore = Pharmacophore.Pharmacophore(activeFeats)
pcophore.setLowerBound(0, 1, 2.251)
pcophore.setUpperBound(0, 1, 2.451)
pcophore.setUpperBound2D(0, 1, 3)
pcophore.setLowerBound(0, 2, 4.970)
pcophore.setUpperBound(0, 2, 5.170)
pcophore.setUpperBound2D(0, 2, 6)
pcophore.setLowerBound(1, 2, 2.681)
pcophore.setUpperBound(1, 2, 2.881)
pcophore.setUpperBound2D(1, 2, 6)
inF = gzip.open(os.path.join(self.dataDir, 'cdk2-syn-clip100.pkl.gz'), 'rb')
nDone = 0
nMatches = 0
nHits = 0
while 1:
try:
name, molPkl, boundsMat = cPickle.load(inF, encoding='latin1')
if PY3:
molPkl = bytes(molPkl, encoding='latin1')
except Exception:
break
nDone += 1
mol = Chem.Mol(molPkl)
boundsMat = rdDistGeom.GetMoleculeBoundsMatrix(mol)
DG.DoTriangleSmoothing(boundsMat)
canMatch, matches = EmbedLib.MatchPharmacophoreToMol(mol, featFactory, pcophore)
if canMatch:
nMatches += 1
r = EmbedLib.MatchPharmacophore(matches, boundsMat, pcophore, useDownsampling=True,
use2DLimits=True, mol=mol)
failed, bm, match, details = r
if not failed:
nHits += 1
self.assertEqual(nDone, 100)
self.assertEqual(nMatches, 93)
#print 'nhits:',nHits
self.assertEqual(nHits, 67)
def _matchMol(self, tpl, pcophore, featFactory, downSample):
name, molPkl, boundsMat = tpl
mol = Chem.Mol(molPkl)
matched, matches = EmbedLib.MatchPharmacophoreToMol(mol, featFactory, pcophore)
if matched:
r = EmbedLib.MatchPharmacophore(matches, boundsMat, pcophore, useDownsampling=downSample)
if r[0]:
return 0
else:
return 1
else:
return 0
def optimize(self, rdmol, boundsMatrix=None, atomMatch=None):
"""
Optimizes the rdmol object using UFF.
Determines the energy level for each of the conformers identified in rdmol.GetConformer.
:param rdmol:
:param boundsMatrix:
:param atomMatch:
:return rdmol, minEid (index of the lowest energy conformer)
"""
energy = 0.0
minEid = 0
lowestE = 9.999999e99 # start with a very high number, which would never be reached
for conf in rdmol.GetConformers():
if (boundsMatrix is None) or (atomMatch is None):
AllChem.UFFOptimizeMolecule(rdmol, confId=conf.GetId())
energy = AllChem.UFFGetMoleculeForceField(
rdmol, confId=conf.GetId()).CalcEnergy()
else:
_, energy = EmbedLib.OptimizeMol(rdmol,
boundsMatrix,
atomMatches=atomMatch,
forceConstant=100000.0)
if energy < lowestE:
minEid = conf.GetId()
lowestE = energy
return rdmol, minEid
def check_mol(mol):
res = []
mol.RemoveAllConformers()
match, mList = EmbedLib.MatchPharmacophoreToMol(mol, feat_fact, pcophore)
#mList = [ m for m in Set( mList ) ]
if match:
num_match = len(mList)
for i in range(num_match):
num_feature = len(mList[i])
for j in range(num_feature):
print mList[i][j].GetAtomIds(), mList[i][j].GetFamily()
bounds = rdDistGeom.GetMoleculeBoundsMatrix(mol)
pList = EmbedLib.GetAllPharmacophoreMatches(mList, bounds, pcophore)
num_match = len(pList)
phMatches = []
for i in range(num_match):
num_feature = len(pList[i])
phMatch = []
for j in range(num_feature):
phMatch.append(pList[i][j].GetAtomIds())
phMatches.append(phMatch)
for phMatch in phMatches:
bm, embeds, nFail = EmbedLib.EmbedPharmacophore(mol,
phMatch,
pcophore,
count=5,
silent=1)
print "-----> embeds num:", len(embeds)
for embed in embeds:
AllChem.UFFOptimizeMolecule(embed)
feats = feat_fact.GetFeaturesForMol(embed)
feats_dict = GetFeatsPerAtoms(feats)
match_feats = [feats_dict[atomid] for atomid in phMatch]
pro_mat = [list(feat.GetPos()) for feat in match_feats]
align_data = rdAlignment.GetAlignmentTransform(
ref_mat, pro_mat, maxIterations=200)
AllChem.TransformMol(embed, align_data[1])
print align_data[0]
print align_data[1]
res.append(embed)
else:
print "no hits"
return res
def testIssue268(self):
from rdkit import RDLogger
#RDLogger.EnableLog('rdApp.debug')
featFactory = ChemicalFeatures.BuildFeatureFactory(os.path.join(self.dataDir, 'Issue268.fdef'))
m1 = Chem.MolFromMolFile(os.path.join(self.dataDir, 'Issue268_Mol1.mol'))
m2 = Chem.MolFromMolFile(os.path.join(self.dataDir, 'Issue268_Mol2.mol'))
with open(os.path.join(self.dataDir, 'Issue268_Pcop.pkl'), 'r') as inTF:
buf = inTF.read().replace('\r\n', '\n').encode('utf-8')
inTF.close()
with io.BytesIO(buf) as inF:
pcop = cPickle.load(inF, encoding='latin1')
#pcop._boundsMat=numpy.array(pcop._boundsMat)
#pcop._boundsMat2D=numpy.array(pcop._boundsMat2D)
#cPickle.dump(pcop,file(os.path.join(self.dataDir,
# 'Issue268_Pcop.new.pkl'),'wb+'))
match, mList1 = EmbedLib.MatchFeatsToMol(m1, featFactory, pcop.getFeatures())
match, mList2 = EmbedLib.MatchFeatsToMol(m2, featFactory, pcop.getFeatures())
b1 = rdDistGeom.GetMoleculeBoundsMatrix(m1)
b2 = rdDistGeom.GetMoleculeBoundsMatrix(m2)
self.assertEqual(len(EmbedLib.MatchPharmacophore(mList1, b1, pcop)[2]), 4)
self.assertEqual(len(EmbedLib.MatchPharmacophore(mList2, b2, pcop)[2]), 4)
self.assertEqual(
len(EmbedLib.MatchPharmacophore(mList1, b1, pcop, mol=m1, use2DLimits=True)[2]), 4)
self.assertEqual(
len(EmbedLib.MatchPharmacophore(mList2, b2, pcop, mol=m2, use2DLimits=True)[2]), 4)
from rdkit import DistanceGeometry as DG
self.assertTrue(DG.DoTriangleSmoothing(b1))
self.assertTrue(DG.DoTriangleSmoothing(b2))
self.assertEqual(len(EmbedLib.MatchPharmacophore(mList1, b1, pcop)[2]), 4)
self.assertEqual(len(EmbedLib.MatchPharmacophore(mList2, b2, pcop)[2]), 4)
self.assertEqual(
len(EmbedLib.MatchPharmacophore(mList1, b1, pcop, mol=m1, use2DLimits=True)[2]), 4)
self.assertEqual(
len(EmbedLib.MatchPharmacophore(mList2, b2, pcop, mol=m2, use2DLimits=True)[2]), 4)
def check_mol( mols, FeatFact, pcophore ): # define FeatFact To Do matched_mols = [] for mol in mols: match, mList = EmbedLib.MatchPharmacophoreToMol( mol, FeatFact, pcophore ) if match: matched_mols.append( mol ) num_match = len( mList ) for i in range( num_match ): num_feature = len( mList[i] ) for j in range( num_feature ): print mList[i][j].GetAtomIds() print len( matched_mols ) return matched_mols
def rd_embed(self):
"""
This portion of the script is literally taken from rmgpy but hacked to work without defining a geometry object
Embed the RDKit molecule and create the crude molecule file.
"""
numConfAttempts = 10000
if (self.bm is None) or (self.atom_match is None):
AllChem.EmbedMultipleConfs(self.rdkit_molecule,
numConfAttempts,
randomSeed=1)
self.rdkit_molecule, minEid = self.optimize_rdkit_molecule()
else:
"""
Embed the molecule according to the bounds matrix. Built to handle possible failures
of some of the embedding attempts.
"""
self.rdkit_molecule.RemoveAllConformers()
for i in range(0, numConfAttempts):
try:
EmbedLib.EmbedMol(self.rdkit_molecule,
self.bm,
atomMatch=self.atom_match)
break
except ValueError:
logging.info(
"RDKit failed to embed on attempt {0} of {1}".format(
i + 1, numConfAttempts))
except RuntimeError:
logging.info("RDKit failed to embed.")
else:
logging.error("RDKit failed all attempts to embed")
return None, None
"""
RDKit currently embeds the conformers and sets the id as 0, so even though multiple
conformers have been generated, only 1 can be called. Below the id's are resolved.
"""
for i in range(len(self.rdkit_molecule.GetConformers())):
self.rdkit_molecule.GetConformers()[i].SetId(i)
self.rdkit_molecule, minEid = self.optimize_rdkit_molecule()
return self.rdkit_molecule, minEid
def _align_molecule(mol, pharmacophore, matches, featFactory, sort=False):
""" Align a molecule to a given pharmacophore.
Uses rdkit alignment algorithm
Parameters
----------
mol : rdkit.Chem.Mol
Molecule to align.
matches : list
If a moleculed is matched to the pharmacophore it will be appended to this list.
pharmacophore : rdkit.Chem.Pharm3D.Pharmacophore
An rdkit pharmacophore
featFactory : rdkit.Chem.rdMolChemicalFeatures.MolChemicalFeatureFactory
The feature factory.
sort : bool, default=False
Whether to sort the list with the matches
"""
bounds_matrix = rdDistGeom.GetMoleculeBoundsMatrix(mol)
# Check if the molecule features can match with the pharmacophore.
can_match, all_matches = EmbedLib.MatchPharmacophoreToMol(mol, featFactory, pharmacophore)
# all_matches is a list of tuples where each tuple contains the chemical features
if can_match:
# Match the molecule to the pharmacophore without aligning it
failed, bounds_matrix_matched, matched_mols, match_details = EmbedLib.MatchPharmacophore(all_matches,
bounds_matrix,
pharmacophore,
useDownsampling=True)
if failed:
return
else:
return
atom_match = [list(x.GetAtomIds()) for x in matched_mols]
try:
mol_H = Chem.AddHs(mol)
# Embed molecule onto the pharmacophore
# embeddings is a list of molecules with a single conformer
b_matrix, embeddings, num_fail = EmbedLib.EmbedPharmacophore(mol_H, atom_match, pharmacophore, count=10)
except:
return
# Align embeddings to the pharmacophore
SSDs = transform_embeddings(pharmacophore, embeddings, atom_match)
if len(SSDs) == 0:
return
best_fit_index = min(enumerate(SSDs), key=itemgetter(1))[0]
try:
mol_id = mol.GetProp("_Name")
except:
mol_id = None
matched_mol = Match(SSDs[best_fit_index], mol_id, embeddings[best_fit_index])
if sort:
# Append to list in ordered manner
try:
# Case when a molecule is repeated. It will throw an error since bisect
# cannot compare molecules.
bisect.insort(matches, matched_mol)
except:
return
else:
matches.append(matched_mol)
