def find_equivelant_torsions(mapped_mol, restricted=False, central_bonds=None):
"""
Final all torsions around a given central bond
Parameters
----------
mapped_mol: oemol. Must contaion map indices
restricted: bool, optional, default False
If True, will also find restricted torsions
central_bonds: list of tuple of ints, optional, defualt None
If provides, only torsions around those central bonds will be given. If None, all torsions in molecule will be found
Returns
-------
eq_torsions: dict
maps central bond to all equivelant torisons
"""
#ToDo check that mol has mapping
from openeye import oechem
mol = oechem.OEMol(mapped_mol)
if not has_atom_map(mol):
raise ValueError("OEMol must have map indices")
terminal_smarts = '[*]~[*]-[X2H1,X3H2,X4H3]-[#1]'
terminal_torsions = _find_torsions_from_smarts(mol, terminal_smarts)
mid_torsions = [[tor.a, tor.b, tor.c, tor.d] for tor in oechem.OEGetTorsions(mapped_mol)]
all_torsions = terminal_torsions + mid_torsions
if restricted:
restricted_smarts = '[*]~[C,c]=,@[C,c]~[*]'
restricted_torsions = _find_torsions_from_smarts(mol, restricted_smarts)
all_torsions = all_torsions + restricted_torsions
tor_idx = []
for tor in all_torsions:
tor_name = (tor[0].GetMapIdx()-1, tor[1].GetMapIdx()-1, tor[2].GetMapIdx()-1, tor[3].GetMapIdx()-1)
tor_idx.append(tor_name)
if central_bonds:
if not isinstance(central_bonds, list):
central_bonds = [central_bonds]
if not central_bonds:
central_bonds = set((tor[1], tor[2]) for tor in tor_idx)
eq_torsions = {cb : [tor for tor in tor_idx if cb == (tor[1], tor[2]) or cb ==(tor[2], tor[1])] for cb in
central_bonds}
return eq_torsions
def GetAdjacentTorsions(mol, refTorsion):
'''
Returns all torsions that are 0 or 1 path length away from
the reference torsion
@param mol: OEGraphMol
@param refTorsion: OETorsion
@return: int
'''
adjTorsions = []
PATH_LENGTH_THRESHOLD = 1
torset = {
str(refTorsion.b.GetIdx()) + "_" + str(refTorsion.c.GetIdx()): True
}
torset[str(refTorsion.c.GetIdx()) + "_" +
str(refTorsion.b.GetIdx())] = True
pred = oechem.OEAndBond(oechem.OEHasOrder(1),
oechem.OENotBond(oechem.OEBondIsInRing()))
for adjTorsion in oechem.OEGetTorsions(mol, pred):
# skip nitrile
order_ab = adjTorsion.a.GetBond(adjTorsion.b).GetOrder()
order_cd = adjTorsion.c.GetBond(adjTorsion.d).GetOrder()
if order_ab == 3 or order_cd == 3:
continue
# skip torsions involving terminal -N-H
if adjTorsion.a.IsHydrogen() and adjTorsion.b.IsNitrogen():
continue
if adjTorsion.d.IsHydrogen() and adjTorsion.c.IsNitrogen():
continue
key1 = str(adjTorsion.b.GetIdx()) + "_" + str(
adjTorsion.c.GetIdx())
key2 = str(adjTorsion.c.GetIdx()) + "_" + str(
adjTorsion.b.GetIdx())
if key1 in torset or key2 in torset:
continue
pathLen = TorsionGenerator.GetMinPathLength(refTorsion, adjTorsion)
if pathLen <= PATH_LENGTH_THRESHOLD:
adjTorsions.append(adjTorsion)
torset[key1] = True
torset[key2] = True
return adjTorsions
def get_torsion(mol, rot_bond):
"""
Find torsion in fragment that corresponds to same rotatable bond in parent molecule
parameters:
----------
mol: OEMol with map indices to parent molecule
rot_bond: tuple
(mapidx_1, mapidx_2)
return:
-------
dihdral: list of dihedral atom indices
"""
tors = [[tor.a, tor.b, tor.c, tor.d] for tor in oechem.OEGetTorsions(mol)]
filtered_torsions = torsions.one_torsion_per_rotatable_bond(tors)
mapped_tors = [[i.GetMapIdx() for i in t] for t in filtered_torsions]
cbs = [(t[1], t[2]) for t in mapped_tors]
try:
dihedral = mapped_tors[cbs.index(rot_bond)]
except ValueError:
dihedral = mapped_tors[cbs.index(tuple(reversed(rot_bond)))]
dihedral = [d - 1 for d in dihedral]
return dihedral
def find_torsion_around_bond(molecule, bond):
"""
Find the torsion around a given bond
Parameters
----------
molecule : molecule with atom maps
bond : tuple of map idx of bond atoms
Returns
-------
list of 4 atom map idx (-1)
Note:
This returns the map indices of the torsion -1, not the atom indices.
"""
from openeye import oechem
if not has_atom_map(molecule):
raise ValueError("Molecule must have atom maps")
#torsions = [[tor.a, tor.b, tor.c, tor.d ] for tor in oechem.OEGetTorsions(molecule)]
terminal_smarts = '[*]~[*]-[X2H1,X3H2,X4H3]-[#1]'
terminal_torsions = _find_torsions_from_smarts(molecule, terminal_smarts)
mid_torsions = [[tor.a, tor.b, tor.c, tor.d] for tor in oechem.OEGetTorsions(molecule)]
all_torsions = terminal_torsions + mid_torsions
tors = one_torsion_per_rotatable_bond(all_torsions)
tor_idx = [tuple(i.GetMapIdx() for i in tor) for tor in tors]
central_bonds = [(tor[1], tor[2]) for tor in tor_idx]
try:
idx = central_bonds.index(bond)
except ValueError:
idx = central_bonds.index(tuple(reversed(bond)))
torsion = [i-1 for i in tor_idx[idx]]
return torsion
def find_torsions(molecule, restricted=True, terminal=True):
#ToDo: Get rid of equivalent torsions. Ex H-C-C-C and C-C-C-H.
"""
This function takes an OEMol (atoms must be tagged with index map) and finds the map indices for torsion that need
to be driven.
Parameters
----------
molecule : OEMol
The atoms in the molecule need to be tagged with map indices
restricted: bool, optional, default True
If True, will find restricted torsions such as torsions in rings and double bonds.
terminal: bool, optional, default True
If True, will find terminal torsions
Returns
-------
needed_torsion_scans: dict
a dictionary that maps internal, terminal and restricted torsions to map indices of torsion atoms
"""
# Check if molecule has map
from openeye import oechem
is_mapped = has_atom_map(molecule)
if not is_mapped:
utils.logger().warning('Molecule does not have atom map. A new map will be generated. You might need a new tagged SMARTS if the ordering was changed')
tagged_smiles = mol_to_smiles(molecule, isomeric=True, mapped=True, explicit_hydrogen=True)
# Generate new molecule with tags
molecule = chemi.smiles_to_oemol(tagged_smiles)
utils.logger().warning('If you already have a tagged SMARTS, compare it with the new one to ensure the ordering did not change')
utils.logger().warning('The new tagged SMARTS is: {}'.format(tagged_smiles))
# ToDo: save the new tagged SMILES somewhere. Maybe return it?
needed_torsion_scans = {'internal': {}, 'terminal': {}, 'restricted': {}}
mol = oechem.OEMol(molecule)
if restricted:
smarts = '[*]~[C,c]=,@[C,c]~[*]' # This should capture double bonds (not capturing rings because OpenEye does not
# generate skewed conformations. ToDo: use scan in geometric or something else to get this done.
restricted_tors = _find_torsions_from_smarts(molecule=mol, smarts=smarts)
if len(restricted_tors) > 0:
restricted_tors_min = one_torsion_per_rotatable_bond(restricted_tors)
for i, tor in enumerate(restricted_tors_min):
tor_name = ((tor[0].GetMapIdx() - 1), (tor[1].GetMapIdx() - 1), (tor[2].GetMapIdx() - 1), (tor[3].GetMapIdx() - 1))
needed_torsion_scans['restricted']['torsion_{}'.format(str(i))] = tor_name
if terminal:
smarts = '[*]~[*]-[X2H1,X3H2,X4H3]-[#1]' # This smarts should match terminal torsions such as -CH3, -NH2, -NH3+, -OH, and -SH
h_tors = _find_torsions_from_smarts(molecule=mol, smarts=smarts)
if len(h_tors) > 0:
h_tors_min = one_torsion_per_rotatable_bond(h_tors)
for i, tor in enumerate(h_tors_min):
tor_name = ((tor[0].GetMapIdx() -1 ), (tor[1].GetMapIdx() - 1), (tor[2].GetMapIdx() - 1), (tor[3].GetMapIdx() - 1))
needed_torsion_scans['terminal']['torsion_{}'.format(str(i))] = tor_name
mid_tors = [[tor.a, tor.b, tor.c, tor.d ] for tor in oechem.OEGetTorsions(mol)]
if mid_tors:
mid_tors_min = one_torsion_per_rotatable_bond(mid_tors)
for i, tor in enumerate(mid_tors_min):
tor_name = ((tor[0].GetMapIdx() - 1), (tor[1].GetMapIdx() - 1), (tor[2].GetMapIdx() - 1), (tor[3].GetMapIdx() - 1))
needed_torsion_scans['internal']['torsion_{}'.format(str(i))] = tor_name
# Check that there are no duplicate torsions in mid and h_torsions
list_tor = list(needed_torsion_scans['internal'].values()) + list(needed_torsion_scans['terminal'].values())
set_tor = set(list_tor)
if not len(set_tor) == len(list_tor):
raise Warning("There is a torsion defined in both mid and terminal torsions. This should not happen. Check "
"your molecule and the atom mapping")
return needed_torsion_scans
def get_canonical_torsions(mol):
'''
Return unique torsions in canonical order.
Only one torsion containing the same central two atoms are return
Cannonical ordering is determined using the order of atoms
in canonical smiles representation
1. generate a canonical smiles representation from the input molecule
2. create a list of (min(b_idx, c_idx), min(a_idx, d_idx), max(a_idx, d_idx), OETorsion)
3. sort the list in #2, extract subset with unique rotatable bonds
:param mol: OEGraphMol
:return: list[OEGraphMol]
'''
CANONICAL_IDX_TAG = 'can_idx'
def assign_canonical_idx(mol):
for atom in mol.GetAtoms():
atom.SetMapIdx(0)
for map_idx, atom in enumerate(mol.GetAtoms(oechem.OEIsHeavy())):
atom.SetMapIdx(map_idx + 1)
can_smiles = oechem.OEMolToSmiles(mol)
can_mol = oechem.OEGraphMol()
# smiles_opt = OEParseSmilesOptions(canon=True)
# OEParseSmiles(can_mol, can_smiles, smiles_opt)
oechem.OESmilesToMol(can_mol, can_smiles)
for can_atom in can_mol.GetAtoms(oechem.OEIsHeavy()):
atom = mol.GetAtom(oechem.OEHasMapIdx(can_atom.GetMapIdx()))
atom.SetData(CANONICAL_IDX_TAG, can_atom.GetIdx())
try:
assign_canonical_idx(mol)
except Exception as e:
print('Error GetCanonicalizedTorsions. ', e)
return None
torsions = []
for torsion in oechem.OEGetTorsions(mol, oechem.OEIsRotor()):
if torsion.a.IsHydrogen() or torsion.b.IsHydrogen() or \
torsion.c.IsHydrogen() or torsion.d.IsHydrogen():
continue
sum_bc = torsion.b.GetData(CANONICAL_IDX_TAG) + torsion.c.GetData(
CANONICAL_IDX_TAG)
min_bc = min(torsion.b.GetData(CANONICAL_IDX_TAG),
torsion.c.GetData(CANONICAL_IDX_TAG))
max_bc = max(torsion.b.GetData(CANONICAL_IDX_TAG),
torsion.c.GetData(CANONICAL_IDX_TAG))
min_ad = min(torsion.a.GetData(CANONICAL_IDX_TAG),
torsion.d.GetData(CANONICAL_IDX_TAG))
max_ad = max(torsion.a.GetData(CANONICAL_IDX_TAG),
torsion.d.GetData(CANONICAL_IDX_TAG))
torsions.append((sum_bc, min_bc, max_bc, min_ad, max_ad, torsion))
# sort
torsions.sort(key=operator.itemgetter(0, 1, 2, 3, 4))
seen = {}
unique_torsions = []
for _, _, _, _, _, torsion in torsions:
bond = mol.GetBond(torsion.b, torsion.c)
if bond is not None and bond.GetIdx() not in seen:
unique_torsions.append(torsion)
seen[bond.GetIdx()] = True
# revert mol to original state
for atom in mol.GetAtoms(oechem.OEIsHeavy()):
atom.SetMapIdx(0)
atom.DeleteData(CANONICAL_IDX_TAG)
return unique_torsions
def GetTorsions(mol):
'''
Goes through each rotatable bond in the molecule
and extracts torsion atoms (a-b-c-d)
Core torsion atoms are extended by one bond
If core or extended atoms are part of a ring,
then entire ring is kept
Keep ortho substitution
Keep functional groups that have at least one atom overlap
with the core/extended torsion atoms
Functional group inclusion criteria:
- <= 5 heavy atoms
- must contain at least one hetero atom
- non-ring
Add methyl cap if bond involving hetero atom is broken
@param mol: OEGraphMol
@type mol: OEGraphMol
@return: list[OEGraphMol]
'''
# mol = OEGraphMol(input_mol)
oechem.OEAssignHybridization(mol)
funcGrps = TorsionGenerator.GetFuncGroups(mol)
includedTorsions = oechem.OEAtomBondSet()
torsionMols = []
for atom in mol.GetAtoms():
atom.SetData("idx", atom.GetIdx() + 1)
torsions = get_canonical_torsions(mol)
if torsions is None:
torsions = oechem.OEGetTorsions(mol, oechem.OEIsRotor())
for torsion in torsions:
if torsion.a.IsHydrogen() or torsion.b.IsHydrogen() or \
torsion.c.IsHydrogen() or torsion.d.IsHydrogen():
continue
torsion_bond = mol.GetBond(torsion.b, torsion.c)
if includedTorsions.HasBond(torsion_bond):
continue
# if includedTorsions.HasAtom(torsion.b) and \
# includedTorsions.HasAtom(torsion.c):
# continue
# revert map idx to zero in original mol
for atom in mol.GetAtoms():
atom.SetMapIdx(0)
# includedTorsions.AddAtom(torsion.b)
# includedTorsions.AddAtom(torsion.c)
includedTorsions.AddBond(torsion_bond)
torsionSet = oechem.OEAtomBondSet(mol.GetBonds())
torsionSet.AddAtoms([torsion.a, torsion.b, torsion.c, torsion.d])
for atom in torsionSet.GetAtoms():
atom.SetMapIdx(1)
# extend core torsion atoms by one bond
nbrs = TorsionGenerator.GetNbrs(torsionSet)
torsionSet.AddAtoms(nbrs)
# include ring atoms
ringAtoms = TorsionGenerator.GetSameRingAtoms(mol, torsionSet)
torsionSet.AddAtoms(ringAtoms)
for atom in torsionSet.GetAtoms():
if not atom.GetMapIdx() == 1:
atom.SetMapIdx(2)
# add functional groups that overlap with torsion set
TorsionGenerator.AddFuncGroupAtoms(funcGrps, torsionSet)
# add relevant ring atoms (ortho substituents and ring H)
TorsionGenerator.AddRelevantRingAtoms(mol, torsion, torsionSet)
# special treatment for C=O
for atom in torsionSet.GetAtoms(
oechem.OEAndAtom(
oechem.OEIsOxygen(),
oechem.OEIsAtomHybridization(
oechem.OEHybridization_sp2))):
for nbr in atom.GetAtoms():
if torsionSet.HasAtom(nbr):
for nbr2 in nbr.GetAtoms(oechem.OEIsHeavy()):
if not torsionSet.HasAtom(nbr2):
nbr2.SetMapIdx(2)
torsionSet.AddAtom(nbr2)
# mark bridging atom and cap if needed
BRIDGE_ATOM_IDX = 4
TorsionGenerator.MarkBridgingAtoms(BRIDGE_ATOM_IDX, mol,
torsionSet)
A_IDX = 11
B_IDX = 12
C_IDX = 13
D_IDX = 14
torsion.a.SetMapIdx(A_IDX)
torsion.b.SetMapIdx(B_IDX)
torsion.c.SetMapIdx(C_IDX)
torsion.d.SetMapIdx(D_IDX)
torsionMol = oechem.OEGraphMol()
oechem.OESubsetMol(torsionMol, mol, torsionSet, True)
torsionMol.Sweep()
torsionMols.append(torsionMol)
# change bridge atom to Carbon
for atom in torsionMol.GetAtoms(
oechem.OEHasMapIdx(BRIDGE_ATOM_IDX)):
atom.SetAtomicNum(oechem.OEElemNo_C)
explicit_valence = atom.GetExplicitValence()
if explicit_valence < 4:
atom.SetImplicitHCount(4 - explicit_valence)
TorsionGenerator.SetSDData(A_IDX, B_IDX, C_IDX, D_IDX, torsion,
torsionMol)
# set map idx to zero in torsion mol
for atom in torsionMol.GetAtoms():
atom.SetMapIdx(0)
# revert map idx to zero in original mol
for atom in mol.GetAtoms():
atom.SetMapIdx(0)
return torsionMols
def generate_torsions(inp_mol,
output_path,
interval,
base_name=None,
tar=True):
"""
This function takes a 3D molecule (pdf, mol2 or sd file) and generates structures for a torsion drive on all torsions
in the molecule. This function uses OpenEye
Parameters
----------
mol : OEMol
molecule to generate 1D torsion scans
output_path: str
path to output file directory
interval: int
angle (in degrees) of interval for torsion drive
base_name: str
base name for file. Default is None. If default, use title in OEMol for base name
tar: bool
If true, will compress output
"""
if not base_name:
base_name = inp_mol.GetTitle()
mid_tors = [[tor.a, tor.b, tor.c, tor.d]
for tor in oechem.OEGetTorsions(inp_mol)]
# This smarts should match terminal torsions such as -CH3, -NH2, -NH3+, -OH, and -SH
smarts = '[*]~[*]-[X2H1,X3H2,X4H3]-[#1]'
qmol = oechem.OEQMol()
if not oechem.OEParseSmarts(qmol, smarts):
warnings.warn('OEParseSmarts failed')
ss = oechem.OESubSearch(qmol)
mol = oechem.OEMol(inp_mol)
h_tors = []
oechem.OEPrepareSearch(mol, ss)
unique = True
for match in ss.Match(mol, unique):
tor = []
for ma in match.GetAtoms():
tor.append(ma.target)
h_tors.append(tor)
# Combine middle and terminal torsions
all_tors = mid_tors + h_tors
# Sort all_tors so that it's grouped by central bond
central_bonds = np.zeros((len(all_tors), 3), dtype=int)
for i, tor in enumerate(all_tors):
central_bonds[i][0] = i
central_bonds[i][1] = tor[1].GetIdx()
central_bonds[i][2] = tor[2].GetIdx()
grouped = central_bonds[central_bonds[:, 2].argsort()]
sorted_tors = [all_tors[i] for i in grouped[:, 0]]
# Keep only one torsion per rotatable bond
tors = []
best_tor = [
sorted_tors[0][0], sorted_tors[0][0], sorted_tors[0][0],
sorted_tors[0][0]
]
first_pass = True
for tor in sorted_tors:
logger().info("Idxs: {} {} {} {}".format(tor[0].GetIdx(),
tor[1].GetIdx(),
tor[2].GetIdx(),
tor[3].GetIdx()))
logger().info("Atom Numbers: {} {} {} {}".format(
tor[0].GetAtomicNum(), tor[1].GetAtomicNum(),
tor[2].GetAtomicNum(), tor[3].GetAtomicNum()))
if tor[1].GetIdx() != best_tor[1].GetIdx() or tor[2].GetIdx(
) != best_tor[2].GetIdx():
new_tor = True
if not first_pass:
logger().info("Adding to list: {} {} {} {}".format(
best_tor[0].GetIdx(), best_tor[1].GetIdx(),
best_tor[2].GetIdx(), best_tor[3].GetIdx()))
tors.append(best_tor)
first_pass = False
best_tor = tor
best_tor_order = tor[0].GetAtomicNum() + tor[3].GetAtomicNum()
logger().info(
"new_tor with central bond across atoms: {} {}".format(
tor[1].GetIdx(), tor[2].GetIdx()))
else:
logger().info("Not a new_tor but now with end atoms: {} {}".format(
tor[0].GetIdx(), tor[3].GetIdx()))
tor_order = tor[0].GetAtomicNum() + tor[3].GetAtomicNum()
if tor_order > best_tor_order:
best_tor = tor
best_tor_order = tor_order
logger().info("Adding to list: {} {} {} {}".format(best_tor[0].GetIdx(),
best_tor[1].GetIdx(),
best_tor[2].GetIdx(),
best_tor[3].GetIdx()))
tors.append(best_tor)
logger().info("List of torsion to drive:")
for tor in tors:
logger().info("Idx: {} {} {} {}".format(tor[0].GetIdx(),
tor[1].GetIdx(),
tor[2].GetIdx(),
tor[3].GetIdx()))
logger().info("Atom numbers: {} {} {} {}".format(
tor[0].GetAtomicNum(), tor[1].GetAtomicNum(),
tor[2].GetAtomicNum(), tor[3].GetAtomicNum()))
conf = mol.GetConfs().next()
coords = oechem.OEFloatArray(conf.GetMaxAtomIdx() * 3)
conf.GetCoords(coords)
# Check if coordinates are not zero
values = np.asarray(
[coords.__getitem__(i) == 0 for i in range(coords.__len__())])
if values.all():
# Generate new coordinates.
mol2 = generate_conformers(mol, max_confs=1)
conf = mol2.GetConfs().next()
coords = oechem.OEFloatArray(conf.GetMaxAtomIdx() * 3)
conf.GetCoords(coords)
mol2.DeleteConfs()
mol.DeleteConfs()
for tor in tors:
tor_name = str((tor[0].GetIdx()) + 1) + '_' + str(
(tor[1].GetIdx()) + 1) + '_' + str(
(tor[2].GetIdx()) + 1) + '_' + str((tor[3].GetIdx()) + 1)
folder = os.path.join(output_path, tor_name)
try:
os.makedirs(folder)
except FileExistsError:
logger().info("Overwriting existing directory {}".format(tor_name))
for angle in range(0, 360, interval):
angle_folder = os.path.join(folder, str(angle))
try:
os.mkdir(angle_folder)
except FileExistsError:
logger().info(
"Overwriting existing directory {}".format(tor_name))
newconf = mol.NewConf(coords)
oechem.OESetTorsion(newconf, tor[0], tor[1], tor[2], tor[3],
radians(angle))
pdb = oechem.oemolostream('{}/{}_{}_{}.pdb'.format(
angle_folder, base_name, tor_name, angle))
oechem.OEWritePDBFile(pdb, newconf)
if tar:
# tar archive output
out = tarfile.open('{}.tar.gz'.format(output_path), mode='w:gz')
os.chdir(output_path)
os.chdir('../')
out.add('{}'.format(base_name))
out.close()
