def test_fold(self):
v = ob.vectorUnsignedInt([0x2A, 0x41])
self.assertEqual(len(v), 2)
x = ob.OBFingerprint.FindFingerprint("FP2")
x.Fold(v, 32)
self.assertEqual(len(v), 1)
self.assertEqual(v[0], (0x2A | 0x41))
v = ob.vectorUnsignedInt([0x01, 0x04, 0x20, 0x00])
self.assertEqual(len(v), 4)
x.Fold(v, 64)
self.assertEqual(len(v), 2)
self.assertEqual(v[0], 0x21)
self.assertEqual(v[1], 0x04)
def testOptions(self):
mol = pybel.readstring("smi", "ICBr")
bv = self.createBitVec(4, (1, 3))
ans = ["[I].[Br]", "I.Br", "I*.Br*"]
ans_atomorder = [[1, 3], [1, 3], [1, 3, 2, 2]]
ans_bondorder = [ [], [], [0, 1] ]
for option in range(3):
nmol = ob.OBMol()
atomorder = ob.vectorUnsignedInt()
bondorder = ob.vectorUnsignedInt()
ok = mol.OBMol.CopySubstructure(nmol, bv, None, option, atomorder, bondorder)
self.assertTrue(ok)
self.assertEqual(pybel.Molecule(nmol).write("smi").rstrip(), ans[option])
self.assertEqual(ans_atomorder[option], list(atomorder))
self.assertEqual(ans_bondorder[option], list(bondorder))
def testOptions(self):
mol = pybel.readstring("smi", "ICBr")
bv = self.createBitVec(4, (1, 3))
ans = ["[I].[Br]", "I.Br", "I*.Br*"]
ans_atomorder = [[1, 3], [1, 3], [1, 3, 2, 2]]
ans_bondorder = [ [], [], [0, 1] ]
for option in range(3):
nmol = ob.OBMol()
atomorder = ob.vectorUnsignedInt()
bondorder = ob.vectorUnsignedInt()
ok = mol.OBMol.CopySubstructure(nmol, bv, None, option, atomorder, bondorder)
self.assertTrue(ok)
self.assertEqual(pybel.Molecule(nmol).write("smi").rstrip(), ans[option])
self.assertEqual(ans_atomorder[option], list(atomorder))
self.assertEqual(ans_bondorder[option], list(bondorder))
def test_get_set(self):
v = ob.vectorUnsignedInt([1, 6])
# XXX Why does GetBit need an actual instance?
x = ob.OBFingerprint.FindFingerprint("FP2")
self.assertTrue(x.GetBit(v, 0))
for i in range(1, 32):
self.assertFalse(x.GetBit(v, i), i)
self.assertFalse(x.GetBit(v, 32))
self.assertTrue(x.GetBit(v, 33))
self.assertTrue(x.GetBit(v, 34))
self.assertFalse(x.GetBit(v, 35))
x.SetBit(v, 35)
self.assertTrue(x.GetBit(v, 35))
def get_mrlogP_descriptors(self, smiles: str, moltitle: str):
# Descriptors are morgan, fp4, then USRCAT
mrlogP_descriptor_length = 128 + 128 + 60
obConversion = openbabel.OBConversion()
obConversion.SetInAndOutFormats("smi", "mdl")
ob_mol = openbabel.OBMol()
# Create RDKit and OpenBabel molecules
rdkit_mol = Chem.AddHs(smiles_to_3dmol(smiles, "querymol"))
obConversion.ReadString(ob_mol, smiles)
# Generate Morgan/ECFP4
morgan_fingerprint = AllChem.GetMorganFingerprintAsBitVect(
Chem.RemoveHs(rdkit_mol), 2, 128).ToBitString()
# Generate USRCAT
usrcat_descriptors = GetUSRCAT(rdkit_mol)
# Generate FP4
fp4fp = openbabel.vectorUnsignedInt()
fingerprinter = openbabel.OBFingerprint.FindFingerprint("FP4")
fingerprinter.GetFingerprint(ob_mol, fp4fp)
openbabel.OBFingerprint.Fold(fingerprinter, fp4fp, 128)
logP_descriptors = np.full((mrlogP_descriptor_length), np.nan)
for i, v in enumerate(morgan_fingerprint):
logP_descriptors[i] = float(v)
fp4_p1 = [float(x) for x in list(format(fp4fp[0], '032b'))]
fp4_p2 = [float(x) for x in list(format(fp4fp[1], '032b'))]
fp4_p3 = [float(x) for x in list(format(fp4fp[2], '032b'))]
fp4_p4 = [float(x) for x in list(format(fp4fp[3], '032b'))]
logP_descriptors[128:256] = fp4_p1 + fp4_p2 + fp4_p3 + fp4_p4
for i, v in enumerate(usrcat_descriptors):
logP_descriptors[256 + i] = float(v)
return logP_descriptors
def test_fp_words(self):
mol = parse_smiles("c1ccccc1O.C#N.[Ge].C1CCC1")
def next_highest_power_of_two(n):
i = 8
while i < n:
i *= 2
return i
for (name, nbits, v0, v1) in (
("FP2", 1021, 0, 1),
("FP3", 55, 67108864, 1159170),
("FP4", 307, 2, 0),
# TODO: change my MACCS.txt so it's correct
# then rerun this test and change to the right answer
("MACCS", 166, 2097156, 256),
):
fingerprinter = ob.OBFingerprint.FindFingerprint(name)
v = ob.vectorUnsignedInt()
fingerprinter.GetFingerprint(mol, v)
size = next_highest_power_of_two(nbits) // 32 # bits-per-int
self.assertEqual(len(v), size)
self.assertEqual(v[0], v0, (name, v[0], v0))
self.assertEqual(v[1], v1, (name, v[1], v1))
def test_tanimoto_with_no_set_bits(self):
v1 = ob.vectorUnsignedInt([0, 0, 0, 0])
x = ob.OBFingerprint.FindFingerprint("FP2")
# Again, this is an arbitrary decision by toolkit providers
self.assertEqual(x.Tanimoto(v1, v1), 0.0)
def test_tanimoto_size_mismatch(self):
v1 = ob.vectorUnsignedInt([0x1, 0x6])
v2 = ob.vectorUnsignedInt([1, 2, 0])
x = ob.OBFingerprint.FindFingerprint("FP2")
self.assertEqual(x.Tanimoto(v1, v2), -1.0)
def test_tanimoto(self):
v1 = ob.vectorUnsignedInt([0x1, 0x6])
v2 = ob.vectorUnsignedInt([0x1, 0x7])
x = ob.OBFingerprint.FindFingerprint("FP2")
self.assertEqual(x.Tanimoto(v1, v2), (1 + 2) / (1 + 3 + 0.0))
