def get_sgs(record_dict, n_min, n_max, method="exhaustive"):
if method == "exhaustive":
return Chem.rdmolops.FindAllSubgraphsOfLengthMToN(
record_dict["mol"], n_min, n_max)
elif method == "RECAP":
hierarchy = Recap.RecapDecompose(record_dict["mol"])
sgs = []
for substructure in hierarchy.GetAllChildren().values():
substructure = Chem.DeleteSubstructs(substructure.mol,
Chem.MolFromSmarts('[#0]'))
edge_idxs = get_substructure_bond_idx(substructure,
record_dict["mol"])
if edge_idxs is not None:
sgs.append(edge_idxs)
return subset_sgs_sizes([sgs], n_min, n_max)
elif method == "BRICS":
substructures = BRICS.BRICSDecompose(record_dict["mol"])
sgs = []
for substructure in substructures:
substructure = Chem.DeleteSubstructs(
Chem.MolFromSmiles(substructure), Chem.MolFromSmarts('[#0]'))
edge_idxs = get_substructure_bond_idx(substructure,
record_dict["mol"])
if edge_idxs is not None:
sgs.append(edge_idxs)
return subset_sgs_sizes([sgs], n_min, n_max)
def get_fragments(self):
fragments = None
if False not in [self.is_mol(),
self.is_small(),
self.has_2_rings()]: # 3 requirements fulfilled
fragments = []
# 1st add scf of the fragments
hierarch = Recap.RecapDecompose(self.get_mol())
ks = hierarch.children
for s, obj in ks.items():
m = obj.mol
if (m is None) or (Chem.GetSSSR(m) < 2):
continue
# Fragments into scaffolds conversion
try:
core = MurckoScaffold.GetScaffoldForMol(m)
except ValueError: # scf calculation not possible
continue
smiles_scf = Chem.MolToSmiles(core)
if Chem.MolFromSmiles(smiles_scf) is None:
continue
fragments.append(smiles_scf)
# 2nd add scf of itself
try:
core = MurckoScaffold.GetScaffoldForMol(self.get_mol())
smiles_scf = Chem.MolToSmiles(core)
if Chem.MolFromSmiles(smiles_scf) is not None:
fragments.append(smiles_scf)
except ValueError: # scf calculation not possible
pass
return (fragments)
def __init__(self,
smi,
lib_smiles,
core_smi=None,
lib_file=None,
fam_sub=True):
self.smi = Chem.MolFromSmiles(smi)
if lib_file is None:
self.lib = gen_lib(lib_smiles)
else:
self.lib = load_lib(lib_file)
if not fam_sub:
self.core = Expand.get_stripped_core(self, core_smi)
self.expanded = Expand.add(self)
else:
main_fragments = Recap.RecapDecompose(self.smi).children.keys()
all_fragments = copy.deepcopy(self.lib)
all_fragments.extend(list(main_fragments))
affin_matrix = Expand.build_affin_matrix(self, all_fragments)
families = Expand.build_families(self, all_fragments, affin_matrix)
self.lib = Expand.get_family_map(self, families, main_fragments)
self.expanded = Expand.fam_add(self)
for x in self.expanded:
print(x)
def pair_frags(fname, out, method='Recap', is_mf=True):
smiles = pd.read_table(fname).Smiles.dropna()
pairs = []
for i, smile in enumerate(tqdm(smiles)):
smile = utils.clean_mol(smile)
mol = Chem.MolFromSmiles(smile)
if method == 'recap':
frags = np.array(sorted(Recap.RecapDecompose(mol).GetLeaves().keys()))
else:
frags = BRICS.BRICSDecompose(mol)
frags = np.array(sorted({re.sub(r'\[\d+\*\]', '*', f) for f in frags}))
if len(frags) == 1: continue
du, hy = Chem.MolFromSmiles('*'), Chem.MolFromSmiles('[H]')
subs = np.array([Chem.MolFromSmiles(f) for f in frags])
subs = np.array([Chem.RemoveHs(Chem.ReplaceSubstructs(f, du, hy, replaceAll=True)[0]) for f in subs])
subs = np.array([m for m in subs if m.GetNumAtoms() > 1])
match = np.array([[m.HasSubstructMatch(f) for f in subs] for m in subs])
frags = subs[match.sum(axis=0) == 1]
frags = sorted(frags, key=lambda x:-x.GetNumAtoms())[:voc.n_frags]
frags = [Chem.MolToSmiles(Chem.RemoveHs(f)) for f in frags]
max_comb = len(frags) if is_mf else 1
for ix in range(1, max_comb+1):
combs = combinations(frags, ix)
for comb in combs:
input = '.'.join(comb)
if len(input) > len(smile): continue
if mol.HasSubstructMatch(Chem.MolFromSmarts(input)):
pairs.append([input, smile])
df = pd.DataFrame(pairs, columns=['Frags', 'Smiles'])
df.to_csv(out, sep='\t', index=False)
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 main(name, argv):
if len(argv) != 2:
print_usage(name)
return
with open(argv[0], 'r') as f:
smile = f.readline().split()[0]
mol = Chem.MolFromSmiles(smile)
size = mol.GetNumHeavyAtoms()
hierarch = Recap.RecapDecompose(mol)
children = []
for child in hierarch.GetAllChildren().keys() + [smile]:
new_smiles = child.replace('[*]', '[H]')
new = Chem.MolFromSmiles(new_smiles)
if not new == None:
new_size = new.GetNumHeavyAtoms()
if new_size > 7 and new_size <= 25:
if rdMolDescriptors.CalcNumRotatableBonds(new) <= 5:
children.append(Chem.MolToSmiles(new, isomericSmiles=True))
#children.append(new_smiles)
core_smile = MurckoScaffold.MurckoScaffoldSmilesFromSmiles(
new_smiles, includeChirality=True)
core = Chem.MolFromSmiles(core_smile)
if rdMolDescriptors.CalcNumRotatableBonds(
core) <= 5 and core.GetNumHeavyAtoms() > 7:
children.append(core_smile)
with open(argv[1], 'w') as f:
i = 1
for m in set(children):
if len(m) > 0:
f.write(m + '\t' + str(i) + '\n')
i += 1
def make_feat_data(mol, offset=1):
res = []
check_atom = set()
nohmol = Chem.RemoveHs(mol)
recap_res = Recap.RecapDecompose(nohmol)
leaves = [
key.replace('*', '').replace('()', '')
for key in recap_res.GetLeaves().keys()
]
leaves = [leave.replace('[H]', '') for leave in leaves if leave != '[H]']
leaves = sorted(leaves,
key=lambda x: Chem.MolFromSmarts(x).GetNumAtoms(),
reverse=True)
if len(leaves) == 0:
line = [i for i in range(mol.GetNumAtoms())]
line = [str(n + offset) for n in line]
line = [Chem.MolToSmiles(mol)] + line
return [line]
for leavsmi in leaves:
leav = Chem.MolFromSmarts(leavsmi)
matches = mol.GetSubstructMatches(leav)
for i, match in enumerate(matches):
line = list(match)
if len(check_atom & set(line)) > 0:
continue
check_atom = check_atom | set(line)
for idx in match:
nei = get_neighbor_h(idx, mol)
line += nei
line = [str(j + offset) for j in line]
line = [leavsmi + '_' + str(i)] + line
res.append(line)
return res
def get_frag_list(mol):
hierarch = Recap.RecapDecompose(mol, minFragmentSize=5)
frag_list = list()
for frag_smi in hierarch.GetLeaves().keys():
frag_mol = Chem.MolFromSmiles(frag_smi)
frag_list.append(frag_mol)
return frag_list
def get_stripped_core(self, core_smi):
if core_smi is None:
print('No core provided generating core with Murcko Scaffold')
core = MurckoScaffold.GetScaffoldForMol(self.smi)
if Chem.MolToSmiles(core) == Chem.MolToSmiles(self.smi):
print(
'Murcko Scaffold failed selecting largest fragment as core'
)
hierarch = Recap.RecapDecompose(self.smi).children.keys()
tmp = Chem.MolFromSmiles(max(hierarch, key=len))
else:
core = Chem.MolFromSmiles(core_smi)
tmp = Chem.ReplaceSidechains(self.smi, core)
return tmp
def get_bicyclic(mol):
bicyclic = []
leaves = Recap.RecapDecompose(mol).GetLeaves()
if len(leaves) != 0:
fragments = leaves.keys()
for fragment in fragments:
fragment_obj = Chem.MolFromSmiles(fragment)
sssr = Chem.GetSSSR(fragment_obj)
if sssr >= 2:
scaffold = Chem.MurckoDecompose(fragment_obj)
bicyclic.append(Chem.MolToSmiles(scaffold))
else:
sssr = Chem.GetSSSR(mol)
if sssr >= 2:
scaffold = Chem.MurckoDecompose(mol)
bicyclic.append(Chem.MolToSmiles(scaffold))
return bicyclic
def build_library(in_smile,
frags,
lib,
rules=os.environ["COVALIB"] +
"/Code/Covalentizer/numbered_reaction.re",
linker_lib=False,
linker_smiles=''):
argv = [in_smile, frags, lib]
with open(argv[0], 'r') as f:
smile = f.readline().split()[0]
mol = Chem.MolFromSmiles(smile)
if mol == None:
return
size = mol.GetNumHeavyAtoms()
hierarch = Recap.RecapDecompose(mol)
children = []
for child in hierarch.GetAllChildren().keys() + [smile]:
new_smiles = child.replace('[*]', '[H]')
new = Chem.MolFromSmiles(new_smiles)
if not new == None:
new_size = new.GetNumHeavyAtoms()
if new_size > 7:
if new_size <= 25 and rdMolDescriptors.CalcNumRotatableBonds(
new) <= 5:
children.append(Chem.MolToSmiles(new, isomericSmiles=True))
core_smile = MurckoScaffold.MurckoScaffoldSmilesFromSmiles(
new_smiles, includeChirality=True)
core = Chem.MolFromSmiles(core_smile)
if new_size <= 25 and rdMolDescriptors.CalcNumRotatableBonds(
core) <= 5 and core.GetNumHeavyAtoms() > 7:
children.append(core_smile)
with open(argv[1], 'w') as f:
i = 1
for m in set(children):
if len(m) > 0:
f.write(m + '\t' + str(i) + '\n')
i += 1
if not linker_lib:
multi_react([rules, argv[1], argv[2]])
else:
multi_linkers([rules, argv[1], argv[2]], linker_smiles)
ts.append(t2 - t1)
if tests[10]:
logger.info('Writing: Mol blocks')
t1 = time.time()
for mol in mols:
mb = Chem.MolToMolBlock(mol)
t2 = time.time()
logger.info('Results10: %.2f seconds' % (t2 - t1))
ts.append(t2 - t1)
if tests[11]:
logger.info('RECAP decomposition')
t1 = time.time()
for mol in mols:
d = Recap.RecapDecompose(mol)
t2 = time.time()
logger.info('Results11: %.2f seconds' % (t2 - t1))
ts.append(t2 - t1)
if tests[12]:
logger.info('Generate 2D coords')
t1 = time.time()
for mol in mols:
AllChem.Compute2DCoords(mol)
t2 = time.time()
logger.info('Results12: %.2f seconds' % (t2 - t1))
ts.append(t2 - t1)
if tests[13]:
logger.info('Generate 3D coords')
'/drug_development/studyRdkit/st_rdcit/data/2d.rxn')
ps = rxn.RunReactants(acid, base)
print('the num of reactions', len(ps))
for reaction in ps:
reactions_smi = Chem.MolToSimiles(reaction)
print('this reaction is', reactions_smi)
# 三、把分子切成片段
# 3.1 Recap方法
# Recap方法通过模拟实验室中常用的化学反应，将反应分子分解成功合理的片段
# Recap 方法返回的是类似于node tree结构的数据。
# > 注：分子片段的Smiles前面通常会加一个 * 如 ‘* c1ccccc1’
m = Chem.MolFromSmiles('c1ccccc1OCCOC(=O)CC')
hierarch = Recap.RecapDecompose(m)
type(hierarch)
# 层次结构的原始分子
print('smi=', hierarch.smiles) # smi= CCC(=O)OCCOc1ccccc1
# 每个节点使用smiles键控的字典跟踪其子节点
ks = hierarch.children.keys()
print(sorted(ks))
# ['*C(=O)CC', '*CCOC(=O)CC', '*CCOc1ccccc1', '*OCCOc1ccccc1', '*c1ccccc1']
# # 3.2 BRICS方法
# RDKit 还提供了另一种把分子切成片段的方法——BRICS方法。 BRICS方法主要是根据可合成的的键对分子进行切断，因此其返回的数据结构是来自于该分子的不同分子片段， 虚拟原子（*）是告诉我们是如何切断的。
# 对下图中的分子进行BRICS分解
smi = 'C=CC(=O)N1CCC(CC1)C2CCNC3=C(C(=NN23)C4=CC=C(C=C4)OC5=CC=CC=C5)C(=O)N'
m = Chem.MolFromSmiles(smi)
def gen_lib(smiles):
mol_lib = [Chem.MolFromSmiles(x) for x in smiles]
hierarch = [Recap.RecapDecompose(x).children.keys() for x in mol_lib]
fragments = [j for i in hierarch for j in i]
return fragments