def _place_between(self, mol: Chem.RWMol, a: int, b: int, aromatic=True):
oribond = mol.GetBondBetweenAtoms(a, b)
if oribond is None:
print('FAIL')
return None # fail
elif aromatic:
bt = Chem.BondType.AROMATIC
else:
bt = oribond.GetBondType()
idx = mol.AddAtom(Chem.Atom(6))
neoatom = mol.GetAtomWithIdx(idx)
atom_a = mol.GetAtomWithIdx(a)
atom_b = mol.GetAtomWithIdx(b)
if aromatic:
neoatom.SetIsAromatic(True)
atom_a.SetIsAromatic(True)
atom_b.SetIsAromatic(True)
# prevent constraints
neoatom.SetBoolProp('_Novel', True)
atom_a.SetBoolProp('_Novel', True)
atom_b.SetBoolProp('_Novel', True)
# fix position
conf = mol.GetConformer()
pos_A = conf.GetAtomPosition(a)
pos_B = conf.GetAtomPosition(b)
x = pos_A.x / 2 + pos_B.x / 2
y = pos_A.y / 2 + pos_B.y / 2
z = pos_A.z / 2 + pos_B.z / 2
conf.SetAtomPosition(idx, Point3D(x, y, z))
# fix bonds
mol.RemoveBond(a, b)
mol.AddBond(a, idx, bt)
mol.AddBond(b, idx, bt)
def _prevent_bridge_ring(self, mol: Chem.RWMol, examplar: Tuple[int]):
## This is really
# examplar is ring
ringatoms = self._get_ring_info(mol) #GetRingInfo().AtomRings()
ringatoms = [ring for ring in ringatoms if set(ring).intersection(examplar)]
ring_idx = list(range(len(ringatoms)))
shared_count = {}
for ra, rb in itertools.combinations(ring_idx, r=2):
shared_count[(ra, rb)] = len(set(ringatoms[ra]).intersection(set(ringatoms[rb])))
if len(shared_count) == 0:
return mol
ra, rb = list(shared_count.keys())[0]
shared = list(set(ringatoms[ra]).intersection(ringatoms[rb]))
pairs = [(a, b) for a, b in itertools.combinations(shared, r=2) if mol.GetBondBetweenAtoms(a, b) is not None]
c = Counter([i for pair in pairs for i in pair])
ring_A, ring_B = ringatoms[ra], ringatoms[rb]
small, big = sorted([ring_A, ring_B], key=lambda ring: len(ring))
inners = [i for i in c if c[i] > 1]
x = list(set(shared).difference(inners))
if len(x) != 2:
log.critical(f'This is impossible. {ringatoms} share {shared} with {inners} in the inside and {x} on the edge?')
return mol
a, b = x
if len(big) > 6:
log.warning(f'Removing {len(inners)} bridging atoms and replacing with fused ring')
# bond the vertices
bt = Chem.BondType.SINGLE # ???
if mol.GetBondBetweenAtoms(a, b) is None:
mol.AddBond(a, b, bt)
else:
log.warning('This is really odd! Why is there a bond already??')
# remove the middle atoms.
for i in sorted(inners, reverse=True):
mol.RemoveAtom(i)
else:
log.warning(f'Shriking the smaller ring to change from bridged to fused.')
# get the neighbour in the small atom to a vertex.
neighs = [neigh for neigh in mol.GetAtomWithIdx(a).GetNeighbors() if
neigh.GetIdx() not in shared and neigh.GetIdx() in small]
neigh = sorted(neighs, key=lambda atom: atom.GetSymbol() != 'C')[0]
bt = mol.GetBondBetweenAtoms(a, neigh.GetIdx()).GetBondType()
mol.RemoveBond(a, neigh.GetIdx())
new_neigh = [neigh for neigh in mol.GetAtomWithIdx(a).GetNeighbors() if neigh.GetIdx() in shared][0]
mol.AddBond(neigh.GetIdx(), new_neigh.GetIdx(), bt)
neigh.SetBoolProp('_Novel', True)
new_neigh.SetBoolProp('_Novel', True)
mol.GetAtomWithIdx(a).SetBoolProp('_Novel', True)
return mol
def join_overclose(self, mol: Chem.RWMol, to_check, cutoff=2.2): # was 1.8
"""
Cutoff is adapted to element.
:param mol:
:param to_check: list of atoms indices that need joining (but not to each other)
:param cutoff: CC bond
:return:
"""
pt = Chem.GetPeriodicTable()
dm = Chem.Get3DDistanceMatrix(mol)
for i in to_check:
atom_i = mol.GetAtomWithIdx(i)
for j, atom_j in enumerate(mol.GetAtoms()):
# calculate cutoff if not C-C
if atom_i.GetSymbol() == '*' or atom_j.GetSymbol() == '*':
ij_cutoff = cutoff
elif atom_i.GetSymbol() == 'C' and atom_j.GetSymbol() == 'C':
ij_cutoff = cutoff
else:
ij_cutoff = cutoff - 1.36 + sum([pt.GetRcovalent(atom.GetAtomicNum()) for atom in (atom_i, atom_j)])
# determine if to join
if i == j or j in to_check:
continue
elif dm[i, j] > ij_cutoff:
continue
else:
self._add_bond_if_possible(mol, atom_i, atom_j)
def join_rings(self, mol: Chem.RWMol, cutoff=1.8):
# special case: x0749. bond between two rings
# namely bonds are added to non-ring atoms. so in the case of bonded rings this is required.
rings = self._get_ring_info(mol)
dm = Chem.Get3DDistanceMatrix(mol)
for ringA, ringB in itertools.combinations(rings, 2):
if not self._are_rings_bonded(mol, ringA, ringB):
mini = np.take(dm, ringA, 0)
mini = np.take(mini, ringB, 1)
d = np.nanmin(mini)
if d < cutoff:
p = np.where(mini == d)
f = ringA[int(p[0][0])]
s = ringB[int(p[1][0])]
#mol.AddBond(f, s, Chem.BondType.SINGLE)
self._add_bond_if_possible(mol, mol.GetAtomWithIdx(f), mol.GetAtomWithIdx(s))
def apply(self, mol: RWMol) -> RWMol:
atom1 = get_atom_ind(mol, self.atom_map1)
atom2 = get_atom_ind(mol, self.atom_map2)
if self.bond_type is None: # delete bond
bond = mol.GetBondBetweenAtoms(atom1, atom2)
if bond is not None:
mol.RemoveBond(atom1, atom2)
else:
b_type = rdchem.BondType.values[self.bond_type]
b_stereo = rdchem.BondStereo.values[self.bond_stereo]
bond = mol.GetBondBetweenAtoms(atom1, atom2)
if bond is None: # add new bond
bond_ind = mol.AddBond(atom1, atom2, order=b_type) - 1
bond = mol.GetBondWithIdx(bond_ind)
else: # change an existing bond
bond.SetBondType(b_type)
bond.SetStereo(b_stereo)
bond.SetBoolProp('is_edited', True)
if b_type == BondType.AROMATIC:
bond.SetIsAromatic(True)
mol.GetAtomWithIdx(atom1).SetIsAromatic(True)
mol.GetAtomWithIdx(atom2).SetIsAromatic(True)
return mol
def apply(self, mol: RWMol) -> RWMol:
num_atoms = mol.GetNumAtoms()
if self.detach:
for i, a in enumerate(mol.GetAtoms()):
m = a.GetAtomMapNum()
if m == self.atom_map2:
for bond in a.GetBonds():
mol.RemoveBond(bond.GetBeginAtomIdx(),
bond.GetEndAtomIdx())
mol.RemoveAtom(i)
num_atoms -= 1
break
atom_ind = get_atom_ind(mol, self.atom_map1)
b_type = rdchem.BondType.values[self.bond_type]
b_stereo = rdchem.BondStereo.values[self.bond_stereo]
old_atom = mol.GetAtomWithIdx(atom_ind)
if old_atom.HasProp('in_reactant'):
self.new_a.SetBoolProp('in_reactant',
old_atom.GetBoolProp('in_reactant'))
if old_atom.HasProp('mol_id'):
self.new_a.SetIntProp('mol_id', old_atom.GetIntProp('mol_id'))
mol.AddAtom(self.new_a)
new_atom_ind = num_atoms
bond_ind = mol.AddBond(atom_ind, new_atom_ind, order=b_type) - 1
new_bond = mol.GetBondWithIdx(bond_ind)
new_bond.SetStereo(b_stereo)
new_bond.SetBoolProp('is_edited', True)
return mol
def apply(self, mol: RWMol) -> RWMol:
atom_ind = get_atom_ind(mol, self.atom_map1)
atom = mol.GetAtomWithIdx(atom_ind)
atom.SetFormalCharge(self.formal_charge)
a_chiral = rdchem.ChiralType.values[self.chiral_tag]
atom.SetChiralTag(a_chiral)
atom.SetNumExplicitHs(self.num_explicit_hs)
atom.SetIsAromatic(self.is_aromatic)
atom.SetBoolProp('is_edited', True)
return mol
def to_rdkit_molecule(data: MoleculeContainer):
"""
MoleculeContainer to RDKit molecule object converter
"""
mol = RWMol()
mapping = {}
bonds = data._bonds
for n, a in data.atoms():
ra = Atom(a.atomic_number)
ra.SetAtomMapNum(n)
if a.charge:
ra.SetFormalCharge(a.charge)
if a.isotope:
ra.SetIsotope(a.isotope)
if a.is_radical:
ra.SetNumRadicalElectrons(1)
mapping[n] = mol.AddAtom(ra)
for n, m, b in data.bonds():
mol.AddBond(mapping[n], mapping[m], _bond_map[b.order])
for n in data._atoms_stereo:
ra = mol.GetAtomWithIdx(mapping[n])
env = bonds[n]
s = data._translate_tetrahedron_sign(n, [x for x in mapping if x in env])
ra.SetChiralTag(_chiral_ccw if s else _chiral_cw)
for nm, s in data._cis_trans_stereo.items():
n, m = nm
if m in bonds[n]: # cumulenes unsupported
nn, nm, *_ = data._stereo_cis_trans[nm]
b = mol.GetBondBetweenAtoms(mapping[n], mapping[m])
b.SetStereoAtoms(mapping[nn], mapping[nm])
b.SetStereo(_cis if s else _trans)
conf = Conformer()
for n, a in data.atoms():
conf.SetAtomPosition(mapping[n], (a.x, a.y, 0))
conf.Set3D(False)
mol.AddConformer(conf, assignId=True)
for c in data._conformers:
conf = Conformer()
for n, xyz in c.items():
conf.SetAtomPosition(mapping[n], xyz)
mol.AddConformer(conf, assignId=True)
SanitizeMol(mol)
AssignStereochemistry(mol, flagPossibleStereoCenters=True, force=True)
return mol
def fragment(self, scaffold):
"""Fragment a scaffold into its next set of Murcko fragments.
Parameters
----------
scaffold : scaffoldgraph.core.Scaffold
Child scaffold to be fragmented.
Returns
-------
list
A list of parent scaffolds representing the next hierarchy.
"""
parents = []
rings = scaffold.ring_systems # ring system information
info = scaffold.rings.info
if rings.count == 1:
return []
for rix, ring in enumerate(rings):
edit = RWMol(scaffold.mol)
remove_atoms = set()
for index, atom in zip(ring.aix, ring.atoms):
if info.NumAtomRings(index) == 1 or any(
[not b.IsInRing() for b in atom.GetBonds()]):
if atom.GetDegree() > 2: # Evoke linker collection
collect_linker_atoms(edit.GetAtomWithIdx(index),
remove_atoms)
else:
remove_atoms.add(index)
else:
remove_atoms.add(index)
for aix in sorted(remove_atoms, reverse=True):
edit.RemoveAtom(aix)
for parent in get_scaffold_frags(edit):
if parent.ring_systems.count == len(rings) - 1:
parent.removed_ring_idx = rix
parents.append(parent)
return parents
def fragment(self, scaffold):
"""Fragment a scaffold into its next set of murcko fragments.
This fragmenter will not dissect fused ring systems.
Parameters
----------
scaffold (sg.core.Scaffold): scaffold to be fragmented.
Returns
-------
parents (list): a list of the next scaffold parents.
"""
parents = []
rings = scaffold.ring_systems # ring system information
info = scaffold.rings.info
if rings.count == 1:
return []
for rix, ring in enumerate(rings):
edit = RWMol(scaffold.mol)
remove_atoms = set()
for index, atom in zip(ring.aix, ring.atoms):
if info.NumAtomRings(index) == 1:
if atom.GetDegree() > 2: # Evoke linker collection
collect_linker_atoms(edit.GetAtomWithIdx(index),
remove_atoms)
else:
remove_atoms.add(index)
else:
remove_atoms.add(index)
for aix in sorted(remove_atoms, reverse=True):
edit.RemoveAtom(aix)
for parent in get_scaffold_frags(edit):
if parent.ring_systems.count == len(rings) - 1:
parent.removed_ring_idx = rix
parents.append(parent)
return parents
def _restore_original_bonding(self, mol: Chem.RWMol, rings) -> None:
to_be_waited_for = []
for ring in rings:
for i in range(len(ring['elements'])):
d = self._get_expansion_for_atom(ring, i)
new_i = self._get_new_index(mol, d['ori_i'], search_collapsed=False)
for old_neigh, bond in zip(d['neighbor'], d['bond']):
bt = getattr(Chem.BondType, bond)
try:
new_neigh = self._get_new_index(mol, old_neigh, search_collapsed=False)
present_bond = mol.GetBondBetweenAtoms(new_i, new_neigh)
if present_bond is None:
mol.AddBond(new_i, new_neigh, bt)
elif present_bond.GetBondType().name != bond:
if self._debug_draw:
print(
f'bond between {new_i} {new_neigh} exists already (has {present_bond.GetBondType().name} expected {bt})')
present_bond.SetBondType(bt)
else:
if self._debug_draw:
print(f'bond between {new_i} {new_neigh} exists already ' + \
f'(has {present_bond.GetBondType().name} expected {bt})')
pass
except ValueError:
if self._debug_draw:
print(f"The neighbour {old_neigh} of {d['ori_i']} with {bt} does not yet exist")
to_be_waited_for.append((new_i, old_neigh, bt))
for new_i, old_neigh, bt in to_be_waited_for:
try:
new_neigh = self._get_new_index(mol, old_neigh, name_restriction=mol.GetAtomWithIdx(new_i).GetProp('_ori_name'))
if self._debug_draw:
print(f'{old_neigh} was missing, but has appeared since as {new_neigh}')
if not mol.GetBondBetweenAtoms(new_i, new_neigh):
mol.AddBond(new_i, new_neigh, bt)
except (KeyError, ValueError) as err:
warn(str(err))
def _delete_collapsed(self, mol: Chem.RWMol):
for a in reversed(range(mol.GetNumAtoms())):
if mol.GetAtomWithIdx(a).GetIntProp('_ori_i') == -1:
mol.RemoveAtom(a)
def fragment(self, scaffold):
"""Fragment a scaffold into its next set of Murcko fragments.
Parameters
----------
scaffold : scaffoldgraph.core.Scaffold
Child scaffold to be fragmented.
Returns
-------
list
A list of parent scaffolds representing the next hierarchy.
"""
parents = [] # container for parent scaffolds
rings = scaffold.rings # ring information
for rix, ring in enumerate(rings): # Loop through all rings and remove
edit = RWMol(scaffold.mol) # Editable molecule
# Collect all removable atoms in the molecule
remove_atoms = set()
for index, atom in zip(ring.aix, ring.atoms):
if rings.info.NumAtomRings(index) == 1:
if atom.GetDegree() > 2: # Evoke linker collection
collect_linker_atoms(edit.GetAtomWithIdx(index),
remove_atoms)
else: # Add ring atom to removable set
remove_atoms.add(index)
else: # Atom is shared between multiple rings
correct_atom_props(edit.GetAtomWithIdx(index))
# Collect removable bonds (this needs to be done to prevent the case where when deleting
# a ring two atoms belonging to the same bond are also part of separate other rings.
# This bond must be broken to prevent an incorrect output)
remove_bonds = set()
for bix in {
x
for x in ring.bix if rings.info.NumBondRings(x) == 1
}:
bond = edit.GetBondWithIdx(bix)
b_x, b_y = bond.GetBeginAtomIdx(), bond.GetEndAtomIdx()
if b_x not in remove_atoms and b_y not in remove_atoms:
remove_bonds.add((b_x, b_y))
correct_atom_props(edit.GetAtomWithIdx(b_x))
correct_atom_props(edit.GetAtomWithIdx(b_y))
# Scheme 4 (scaffold tree rule)
if self.use_scheme_4 is not False and len(ring) == 3:
atomic_nums = [a.GetAtomicNum() for a in ring.atoms]
if len([a for a in atomic_nums if a != 1 and a != 6]) == 1:
shared = {
x
for x in ring.bix if rings.info.NumBondRings(x) > 1
}
if len(shared) == 1:
bond = edit.GetBondWithIdx(shared.pop())
bond.SetBondType(BondType.DOUBLE)
# Remove collected atoms and bonds
for bix in remove_bonds:
edit.RemoveBond(*bix)
for aix in sorted(remove_atoms, reverse=True):
edit.RemoveAtom(aix)
# Add new parent scaffolds to parent list
for parent in get_scaffold_frags(edit):
if parent.rings.count == len(rings) - 1:
parent.removed_ring_idx = rix
parents.append(parent)
return parents
def _join_atoms(self,
combo: Chem.RWMol,
anchor_A: int,
anchor_B: int,
distance: float,
linking: bool = True):
"""
extrapolate positions between. by adding linkers if needed.
"""
conf = combo.GetConformer()
pos_A = conf.GetAtomPosition(anchor_A)
pos_B = conf.GetAtomPosition(anchor_B)
n_new = int(round(distance / 1.22) - 1)
xs = np.linspace(pos_A.x, pos_B.x, n_new + 2)[1:-1]
ys = np.linspace(pos_A.y, pos_B.y, n_new + 2)[1:-1]
zs = np.linspace(pos_A.z, pos_B.z, n_new + 2)[1:-1]
# correcting for ring marker atoms
def is_ring_atom(anchor: int) -> bool:
atom = combo.GetAtomWithIdx(anchor)
if atom.HasProp('_ori_i') and atom.GetIntProp('_ori_i') == -1:
return True
else:
return False
if is_ring_atom(anchor_A):
distance -= 1.35 + 0.2 # Arbitrary + 0.2 to compensate for the ring not reaching (out of plane).
n_new -= 1
xs = xs[1:]
ys = ys[1:]
zs = zs[1:]
if is_ring_atom(anchor_B):
distance -= 1.35 + 0.2 # Arbitrary + 0.2 to compensate for the ring not reaching (out of plane).
n_new -= 1
xs = xs[:-1]
ys = ys[:-1]
zs = zs[:-1]
# notify that things could be leary.
if distance < 0:
self.journal.debug(
f'Two ring atoms detected to be close. Joining for now.' +
' They will be bonded/fused/spiro afterwards')
# check if valid.
if distance > self.joining_cutoff:
msg = f'Atoms {anchor_A}+{anchor_B} are {distance} Å away. Cutoff is {self.joining_cutoff}.'
self.journal.warning(msg)
raise ConnectionError(msg)
# place new atoms
self.journal.debug(
f'Molecules will be joined via atoms {anchor_A}+{anchor_B} ({distance} Å) via the addition of {n_new} atoms.'
)
previous = anchor_A
if linking is False and n_new > 0:
self.journal.warning(
f'Was going to bond {anchor_A} and {anchor_B} but reconsidered.'
)
elif linking is True and n_new <= 0:
combo.AddBond(previous, anchor_B, Chem.BondType.SINGLE)
new_bond = combo.GetBondBetweenAtoms(previous, anchor_B)
BondProvenance.set_bond(new_bond, 'main_novel')
elif linking is False and n_new <= 0:
combo.AddBond(previous, anchor_B, Chem.BondType.SINGLE)
new_bond = combo.GetBondBetweenAtoms(previous, anchor_B)
BondProvenance.set_bond(new_bond, 'other_novel')
elif linking is True and n_new > 0:
for i in range(n_new):
# make oxygen the first and last bridging atom.
if i == 0 and combo.GetAtomWithIdx(
anchor_A).GetSymbol() == 'C':
new_atomic = 8
elif i > 2 and i == n_new - 1 and combo.GetAtomWithIdx(
anchor_B).GetSymbol() == 'C':
new_atomic = 8
else:
new_atomic = 6
idx = combo.AddAtom(Chem.Atom(new_atomic))
new = combo.GetAtomWithIdx(idx)
new.SetBoolProp('_Novel', True)
new.SetIntProp('_ori_i', 999)
conf.SetAtomPosition(
idx, Point3D(float(xs[i]), float(ys[i]), float(zs[i])))
combo.AddBond(idx, previous, Chem.BondType.SINGLE)
new_bond = combo.GetBondBetweenAtoms(idx, previous)
BondProvenance.set_bond(new_bond, 'linker')
previous = idx
combo.AddBond(previous, anchor_B, Chem.BondType.SINGLE)
new_bond = combo.GetBondBetweenAtoms(previous, anchor_B)
BondProvenance.set_bond(new_bond, 'linker')
else:
raise ValueError('Impossible')
return combo.GetMol()