def all_join_on_attach_point(mol1, mol2):
"""Join two molecules on all possible attaching point
Arguments
---------
mol1: <Chem.Mol>
input molecule 1
mol2: <Chem.Mol>
input molecule 2
Returns:
iterator of all possible way to attach both molecules from dummy indicators.
"""
atom_map_min = 100
mol_idxs = []
count = 0
mod_mols = []
for ind, m in enumerate([mol1, mol2]):
atms = [(a.GetIdx(), a) for a in m.GetAtoms()
if not a.IsInRing() and a.GetAtomicNum() == 0]
atms.sort(reverse=True, key=operator.itemgetter(0))
for a_idx, a in atms:
for a_nei in a.GetNeighbors():
a_nei.SetAtomMapNum(atom_map_min + count)
count += 1
mod_mol = dm.fix_mol(m)
mod_mols.append(mod_mol)
mol_idxs.append([
a.GetIdx() for a in mod_mol.GetAtoms()
if a.GetAtomMapNum() >= atom_map_min
])
for ind1, ind2 in itertools.product(*mol_idxs):
yield random_fragment_add(copy.copy(mod_mols[0]),
copy.copy(mod_mols[1]), ind1, ind2)
def _preprocess(i, row):
# print('hello')
mol = dm.to_mol(str(row[smiles_column]), ordered=True)
mol = dm.fix_mol(mol)
mol = dm.sanitize_mol(mol, sanifix=True, charge_neutral=False)
mol = dm.standardize_mol(mol, disconnect_metals=False, normalize=True, reionize=True, uncharge=False, stereo=True)
fingerprint_function = rdMolDescriptors.GetMorganFingerprintAsBitVect
pars = { "radius": 2,
"nBits": 8192,
"invariants": [],
"fromAtoms": [],
"useChirality": True,
"useBondTypes": True,
"useFeatures": False,
}
fp = fingerprint_function(mol, **pars)
row["standard_smiles"] = dm.standardize_smiles(dm.to_smiles(mol))
row["selfies"] = dm.to_selfies(mol)
row["inchi"] = dm.to_inchi(mol)
row["inchikey"] = dm.to_inchikey(mol)
row["onbits_fp"] =list(fp.GetOnBits())
return row
def recap(
mol: Chem.Mol,
remove_parent: bool = False,
sanitize: bool = True,
fix: bool = True,
):
"""Fragment the molecule using the recap algorithm.
Args:
mol: a molecule.
remove_parent: Remove parent from the fragments.
sanitize: Wether to sanitize the fragments.
fix: Wether to fix the fragments.
"""
res = Recap.RecapDecompose(mol)
frags = [dm.to_mol(x) for x in res.GetAllChildren().keys()]
if fix:
frags = [dm.fix_mol(x) for x in frags]
if sanitize:
frags = [dm.sanitize_mol(x) for x in frags]
frags = [x for x in frags if x is not None]
if remove_parent:
return frags
return [mol] + frags
def frag(
mol: Chem.Mol,
remove_parent: bool = False,
sanitize: bool = True,
fix: bool = True,
):
"""Generate all possible fragmentation of a molecule.
Args:
mol: a molecule.
remove_parent: Remove parent from the fragments.
sanitize: Wether to sanitize the fragments.
fix: Wether to fix the fragments.
"""
frags = FraggleSim.generate_fraggle_fragmentation(mol)
smiles = set([])
for seq in frags:
smiles |= {s.strip() for s in seq.split(".")}
smiles = list(sorted(smiles, reverse=True))
frags = [dm.to_mol(s) for s in smiles]
if fix:
frags = [dm.fix_mol(x) for x in frags]
if sanitize:
frags = [dm.sanitize_mol(x) for x in frags]
frags = [x for x in frags if x is not None]
if remove_parent:
return frags
return [mol] + frags
def brics(
mol: Chem.Mol,
singlepass: bool = True,
remove_parent: bool = False,
sanitize: bool = True,
fix: bool = True,
):
"""Run BRICS on the molecules and potentially fix dummy atoms.
Args:
mol: a molecule.
singlepass: Single pass for `BRICSDecompose`.
remove_parent: Remove parent from the fragments.
sanitize: Wether to sanitize the fragments.
fix: Wether to fix the fragments.
"""
frags = BRICS.BRICSDecompose(mol, returnMols=True, singlePass=singlepass)
frags = list(frags)
if fix:
frags = [dm.fix_mol(x) for x in frags]
if sanitize:
frags = [dm.sanitize_mol(x) for x in frags]
if remove_parent:
frags.pop(0)
frags = [x for x in frags if x is not None]
return frags
def compute_reaction_product(out, single_output=True):
"""Compute the product of a reaction"""
out = [dm.fix_mol(x[0], n_iter=0) for x in out]
if not single_output:
return [dm.sanitize_mol(x) for x in out]
# Might be a important to make a tradeoff decision in selecting products for greater speed.
# product = sorted(out, key=lambda x: MoleculeEnv.compute_reward_from_mol(x, True))[-1]
# sampling from list of products is an alternative
return dm.sanitize_first(np.random.permutation(out))
def test_fixmol():
sm = "C.Cl.CC.[H][N:1]1(C)=CC(O)=CC2CCCCC12"
mol = Chem.MolFromSmiles(sm, sanitize=False)
# mol.UpdatePropertyCache(False)
# Chem.Kekulize(mol)
res = dm.fix_mol(mol, n_iter=1) # copy by default
# should still be invalid in term of valence for nitrogen
assert not dm.incorrect_valence(res)
res2 = dm.fix_mol(mol, n_iter=2)
# not expecting difference between res2 and res3
assert Chem.MolToSmiles(res) == Chem.MolToSmiles(res2)
# only largest expected_here
res_largest = dm.fix_mol(mol, largest_only=True)
dm.fix_mol(mol, remove_singleton=True, largest_only=True)
assert len(Chem.rdmolops.GetMolFrags(res_largest)) == 1
expected_largest_fix = dm.standardize_smiles("OC1=CC2CCCCC2[N:1]=C1")
assert dm.standardize_smiles(
Chem.MolToSmiles(res_largest)) == expected_largest_fix
res_no_singleton = dm.fix_mol(mol, n_iter=2, remove_singleton=True)
assert len(Chem.rdmolops.GetMolFrags(res_largest)) == 1
assert len(Chem.rdmolops.GetMolFrags(res_no_singleton)) == 2
def _preprocess(i, row):
# print('hello')
try:
mol = dm.to_mol(str(row[smiles_column]), ordered=True)
mol = dm.fix_mol(mol)
mol = dm.sanitize_mol(mol, sanifix=True, charge_neutral=False)
mol = dm.standardize_mol(mol,
disconnect_metals=False,
normalize=True,
reionize=True,
uncharge=False,
stereo=True)
opts = StereoEnumerationOptions(unique=True,
maxIsomers=20,
rand=0xf00d)
isomers = EnumerateStereoisomers(mol, options=opts)
enum_smiles = sorted(
Chem.MolToSmiles(y, isomericSmiles=True) for y in isomers)
smiles_list = []
for count, smi in enumerate(enum_smiles):
smiles_string = smi
smiles_list.append(smiles_string)
# fingerprint_function = rdMolDescriptors.GetMorganFingerprintAsBitVect
# pars = { "radius": 2,
# "nBits": 8192,
# "invariants": [],
# "fromAtoms": [],
# "useChirality": False,
# "useBondTypes": True,
# "useFeatures": False,
# }
# fp = fingerprint_function(mol, **pars)
row["standard_smiles"] = dm.standardize_smiles(dm.to_smiles(mol))
row["selfies"] = dm.to_selfies(mol)
row["inchi"] = dm.to_inchi(mol)
row["inchikey"] = dm.to_inchikey(mol)
row["enumerated_smiles"] = smiles_list
# row["onbits_fp"] =list(fp.GetOnBits())
return row
except ValueError:
row["standard_smiles"] = 'dropped'
row["selfies"] = 'dropped'
row["inchi"] = 'dropped'
row["inchikey"] = 'dropped'
row["enumerated_smiles"] = list('dropped')
return row
def trim_side_chain(mol: Chem.rdchem.Mol, core, unwanted_side_chains):
"""Trim list of side chain from a molecule."""
mol = Chem.AddHs(mol)
match = mol.GetSubstructMatch(core)
map2idx = {}
map2nei = {}
unwanted2map = {}
for patt in unwanted_side_chains:
unwanted2map[patt] = [
a.GetAtomMapNum() for a in patt.GetAtoms() if a.GetAtomMapNum()
]
unwanted_mapping = list(
itertools.chain.from_iterable(unwanted2map.values()))
for atom in core.GetAtoms():
num = atom.GetAtomMapNum()
if num and num in unwanted_mapping:
mol_atom_idx = match[atom.GetIdx()]
map2idx[mol_atom_idx] = num
nei_atoms = mol.GetAtomWithIdx(mol_atom_idx).GetNeighbors()
map2nei[mol_atom_idx] = [
n.GetIdx() for n in nei_atoms if n.GetIdx() in match
]
emol = Chem.EditableMol(mol)
for atom_idx, atom_map in map2idx.items():
dummy = Chem.rdchem.Atom("*")
dummy.SetAtomMapNum(atom_map)
nei_idx = map2nei.get(atom_idx, [None])[0]
if nei_idx:
bond = mol.GetBondBetweenAtoms(atom_idx, nei_idx)
emol.RemoveBond(atom_idx, nei_idx)
new_ind = emol.AddAtom(dummy)
emol.AddBond(nei_idx, new_ind, bond.GetBondType())
mol = emol.GetMol()
mol = Chem.RemoveHs(mol)
query_param = AdjustQueryParameters()
query_param.makeDummiesQueries = False
query_param.adjustDegree = False
query_param.aromatizeIfPossible = True
for patt, _ in unwanted2map.items():
cur_frag = dm.fix_mol(patt)
mol = Chem.DeleteSubstructs(mol, cur_frag, onlyFrags=True)
return dm.keep_largest_fragment(mol)
def break_mol(
mol: Chem.Mol,
minFragmentSize: int = 1,
silent: bool = True,
onlyUseReactions: list = [],
randomize: bool = False,
mode: str = "brics",
returnTree: bool = False,
):
"""Breaks a molecules into a list of fragment."""
if mode.lower() == "brics":
all_reactions = ALL_BRICS
all_reactions_type = ALL_BRICS_TYPE
elif mode.lower() == "rxn":
all_reactions = ALL_RXNS
all_reactions_type = ALL_RXNS_TYPE
else:
all_reactions = ALL_BRICS + ALL_RXNS
all_reactions_type = ALL_BRICS_TYPE + ALL_RXNS_TYPE
if randomize:
p = np.random.permutation(len(all_reactions))
all_reactions = [all_reactions[ind] for ind in p]
all_reactions_type = [all_reactions_type[ind] for ind in p]
nx = dm.graph._get_networkx()
mSmi = Chem.MolToSmiles(mol, isomericSmiles=True)
G = nx.DiGraph()
node_num = 0
G.add_node(node_num, smiles=mSmi, mol=mol)
allNodes = set()
activePool = {mSmi: node_num}
allNodes.add(mSmi)
while activePool:
nSmi = list(activePool.keys())[0]
parent = activePool.pop(nSmi)
node = G.nodes[parent]
mol = node["mol"]
for rxnIdx, reaction in zip(all_reactions_type, all_reactions):
if onlyUseReactions and rxnIdx not in onlyUseReactions:
continue
ps = reaction.RunReactants((mol,))
if ps:
all_pass = [
all([prod.GetNumAtoms(onlyExplicit=True) > minFragmentSize for prod in p_])
for p_ in ps
]
nz_i = 0
while nz_i < len(all_pass) and not all_pass[nz_i]:
nz_i += 1
if not silent:
print(nSmi, "->", len(ps), "products and selected ", nz_i)
# display(MolsToGridImage(list(itertools.chain(*list(ps))), molsPerRow=2))
prodSeq = ps[nz_i % len(all_pass)]
seqOk = True
# we want to disqualify small fragments, so sort the product sequence by size
prodSeq = [(prod.GetNumAtoms(onlyExplicit=True), prod) for prod in prodSeq]
prodSeq.sort(key=lambda x: x[0])
for _, prod in prodSeq:
prod.sanitized = True
try:
Chem.SanitizeMol(prod)
except:
if dm.sanitize_mol(prod) is None:
seqOk = False
break
continue
pSmi = Chem.MolToSmiles(prod, isomericSmiles=True)
seqOk = seqOk and (dm.to_mol(pSmi) is not None)
notDummies = sum([atm.GetSymbol() != "*" for atm in prod.GetAtoms()])
# nDummies = pSmi.count('*')
# if minFragmentSize > 0 and (nats - nDummies < minFragmentSize):
if minFragmentSize > 0 and notDummies < minFragmentSize:
seqOk = False
break
prod.pSmi = pSmi
if seqOk:
for _, prod in prodSeq:
if not prod.sanitized:
continue
pSmi = prod.pSmi
node_num += 1
usmi = Chem.MolToSmiles(dm.fix_mol(prod), isomericSmiles=True)
G.add_node(node_num, smiles=usmi, mol=prod)
G.add_edge(parent, node_num)
if usmi not in allNodes:
activePool[pSmi] = node_num
allNodes.add(usmi)
G.nodes[parent]["rxn"] = rxnIdx
break # at least one reaction matches
leaves_smiles = [
G.nodes[n]["smiles"] for n in G.nodes() if G.in_degree(n) != 0 and G.out_degree(n) == 0
]
if returnTree:
return leaves_smiles, allNodes, G
return leaves_smiles, allNodes
def _preprocess(i, row):
'''Takes a smiles string and generates a clean rdkit mol with datamol. The stereoisomers
are then enumerated while holding defined stereochemistry. Morgan fingerprints are then
generated using RDkit with and without stereochemistry. The try/except logic deals with
RDkit mol failures on conversion of an invalid smiles string. Smarts are added for later
searching.'''
try:
mol = dm.to_mol(str(row[smiles_column]), ordered=True)
mol = dm.fix_mol(mol)
mol = dm.sanitize_mol(mol, sanifix=True, charge_neutral=False)
mol = dm.standardize_mol(mol, disconnect_metals=False, normalize=True, reionize=True, uncharge=False, stereo=True)
opts = StereoEnumerationOptions(unique=True,maxIsomers=20,rand=0xf00d)
isomers = EnumerateStereoisomers(mol, options=opts)
enum_smiles = sorted(Chem.MolToSmiles(y,isomericSmiles=True) for y in isomers)
# enum_dm_smiles = sorted(dm.standardize_smiles(dm.to_smiles(x)) for x in isomers)
smiles_list = []
achiral_fp_lis = []
chiral_fp_lis = []
# standard_smiles_list = []
for count, smi in enumerate(enum_smiles):
smiles_string = smi
mol = dm.to_mol(smi, ordered=True)
mol = dm.fix_mol(mol)
mol = dm.sanitize_mol(mol, sanifix=True, charge_neutral=False)
mol = dm.standardize_mol(mol, disconnect_metals=False, normalize=True, reionize=True, uncharge=False, stereo=True)
fingerprint_function = rdMolDescriptors.GetMorganFingerprintAsBitVect
pars = { "radius": 2,
"nBits": 8192,
"invariants": [],
"fromAtoms": [],
"useChirality": True,
"useBondTypes": True,
"useFeatures": False, }
pars2 = { "radius": 2,
"nBits": 8192,
"invariants": [],
"fromAtoms": [],
"useChirality": False,
"useBondTypes": True,
"useFeatures": False, }
fp = fingerprint_function(mol, **pars)
fp1 = fingerprint_function(mol, **pars2)
smiles_list.append(dm.standardize_smiles(smiles_string))
achiral_fp_lis.append(list(fp1.GetOnBits()))
chiral_fp_lis.append(list(fp.GetOnBits()))
row["standard_smiles"] = dm.standardize_smiles(dm.to_smiles(mol))
row["smarts"] = dm.to_smarts(mol)
row["selfies"] = dm.to_selfies(mol)
row["enumerated_smiles"] = smiles_list
row["achiral_fp"] = achiral_fp_lis
row["chiral_fp"] = chiral_fp_lis
# row["dm_enumerated_smiles"] = enum_dm_smiles_lis
# row["onbits_fp"] =list(fp.GetOnBits())
return row
except ValueError:
# row["standard_smiles"] = 'dropped'
# row["selfies"] = 'dropped'
# row["inchi"] = 'dropped'
# row["inchikey"] = 'dropped'
row["standard_smiles"] = 'dropped'
row["smarts"] = 'dropped'
row["selfies"] = 'dropped'
row["enumerated_smiles"] = list('dropped')
row["achiral_fp"] = list('dropped')
row["chiral_fp"] = list('dropped')
# row["dm_enumerated_smiles"] = 'dropped'
return row
