def split_strip(inchi, strip=True):
mol = rdkit.Chem.MolFromInchi(inchi)
split_inchis = set()
for m in rdkit.Chem.rdmolops.GetMolFrags(mol,
asMols=True): # split
i = rdkit.Chem.MolToInchi(m)
if strip:
c = ic.inchi_conn_layer(i)
if not c in self.conn_split_inactive_inchis:
split_inchis.add(i)
else:
split_inchis.add(i)
return split_inchis
def query_inchi_conn(self, inchi, strip=True):
if strip:
try:
inchi = ic.strip_inchi(
inchi, exclude_inchis=self.conn_split_inactive_inchis
) # filter inchi
except:
pass
inchi_conn = ic.inchi_conn_layer(inchi)
r = set()
if inchi_conn in self.inchi_connectivity_index:
r.update(self.inchi_connectivity_index[inchi_conn])
if inchi_conn in self.inchi_split_connectivity_index:
r.update(self.inchi_split_connectivity_index[inchi_conn])
return r
def inchi_overlap(
self,
inchi1,
inchi2,
strip=True,
consistency=True,
filter_layers={'h', 'f', 'p', 'q', 'i', 't', 'b', 'm', 's'}):
def split_strip(inchi, strip=True):
mol = rdkit.Chem.MolFromInchi(inchi)
split_inchis = set()
for m in rdkit.Chem.rdmolops.GetMolFrags(mol,
asMols=True): # split
i = rdkit.Chem.MolToInchi(m)
if strip:
c = ic.inchi_conn_layer(i)
if not c in self.conn_split_inactive_inchis:
split_inchis.add(i)
else:
split_inchis.add(i)
return split_inchis
if isinstance(inchi1, list):
inchi1 = set(inchi1)
if isinstance(inchi2, list):
inchi2 = set(inchi2)
if isinstance(inchi1, set):
split_inchis1 = set(inchi1)
else:
split_inchis1 = split_strip(inchi1, strip=strip)
if isinstance(inchi2, set):
split_inchis2 = set(inchi2)
else:
split_inchis2 = split_strip(inchi2, strip=strip)
# get conn layer
conn_split_inchis1 = defaultdict(set)
for i in split_inchis1:
conn_split_inchis1[ic.inchi_conn_layer(i)].add(i)
conn_split_inchis2 = defaultdict(set)
for i in split_inchis2:
conn_split_inchis2[ic.inchi_conn_layer(i)].add(i)
# compare conn layers to get candidate matches
candidates = set()
for k in set(conn_split_inchis1.keys()) & set(
conn_split_inchis2.keys()):
candidates.update(
it.product(conn_split_inchis1[k], conn_split_inchis2[k]))
# check these candidate matches by chacking consistency
matches = set()
for i1, i2 in candidates:
if consistency:
c, s = ic.compare_consistent(i1,
i2,
filter_layers=filter_layers)
if c:
matches.add((i1, i2))
else:
if i1 == i2:
matches.add((i1, i2))
# return overlap between the split inchi fragments
i1_matches = {i1 for (i1, i2) in matches}
i2_matches = {i2 for (i1, i2) in matches}
overlap = i1_matches | i2_matches
return split_inchis1 - overlap, matches, split_inchis2 - overlap # remainder from inchi1, paired matches, remainder from inchi2
def load_inactive_compounds(self, data_dir=None):
def trim_whitespace(s):
while True:
if s[-1:] in {'\n', '\r', '\t', ' '}:
s = s[:-1]
else:
break
while True:
if s[:1] in {'\n', '\r', '\t', ' '}:
s = s[1:]
else:
break
return s
if data_dir is None:
data_dir = self.data_dir
# load inactive data
with open(f"{data_dir}/exclude_inchis.json", 'rt') as f:
self.exclude_inchis = set(json.load(f))
with open(f"{data_dir}/split_inactive_inchis.json", 'rt') as f:
self.split_inactive_inchis = set(json.load(f))
with open(f"{data_dir}/salts.smi", 'rt') as f:
cols = ['name', 'smiles']
salts = [{
cols[i]: trim_whitespace(v)
for i, v in enumerate(l.split('\t'))
} for l in f]
with open(f"{data_dir}/solvents.smi", 'rt') as f:
cols = ['name', 'smiles']
solvents = [{
cols[i]: trim_whitespace(v)
for i, v in enumerate(l.split('\t'))
} for l in f]
salts_inchi = set()
for d in salts:
try:
mol = rdkit.Chem.MolFromSmiles(d['smiles'])
inchi = rdkit.Chem.MolToInchi(mol)
salts_inchi.add(inchi)
except Exception as e:
pass
solvents_inchi = set()
for d in solvents:
try:
mol = rdkit.Chem.MolFromSmiles(d['smiles'])
inchi = rdkit.Chem.MolToInchi(mol)
solvents_inchi.add(inchi)
except Exception as e:
pass
self.split_inactive_inchis.update(salts_inchi)
self.split_inactive_inchis.update(solvents_inchi)
self.conn_split_inactive_inchis = {
ic.inchi_conn_layer(i)
for i in self.split_inactive_inchis
}
def fetch_data(self, data_dir=None):
if data_dir is None:
data_dir = self.data_dir
self.load_inactive_compounds(data_dir=data_dir)
chembl_cursor = self.chembl_db.cursor()
# fetch structures
self.compound_inchis = {}
for molregno, inchi in tqdm(chembl_cursor.execute(
"select MOLREGNO, STANDARD_INCHI from CHEMBL.COMPOUND_STRUCTURES"
),
leave=True,
position=0,
desc='ChEMBL structures'):
if inchi is None:
continue
ci = self.chembl_index.get_chembl_ident(molregno=molregno)
self.compound_inchis[ci] = inchi
structure_id2drugbase_id = {}
for drugbase_id, mrn, structure_id in tqdm(chembl_cursor.execute(
'select ID, MOLREGNO, MOLECULE_STRUCTURE_ID from DRUGBASE.MOLECULE_DICTIONARY'
),
leave=True,
position=0,
desc='Drugbase structure IDs'
):
ci = self.chembl_index.get_chembl_ident(drugbase_id=drugbase_id)
structure_id2drugbase_id[structure_id] = ci
for structure_id, inchi in tqdm(chembl_cursor.execute(
"select MOLECULE_STRUCTURE_ID, INCHI from DRUGBASE.MOLECULE_STRUCTURE"
),
leave=True,
position=0,
desc='Drugbase structures'):
if inchi is None:
continue
inchi = inchi.read()
if inchi is None:
continue
if structure_id in structure_id2drugbase_id:
ci = structure_id2drugbase_id[structure_id]
self.compound_inchis[ci] = inchi
self.inchi_index = defaultdict(set)
for k, v in self.compound_inchis.items():
self.inchi_index[v].add(k)
# split InChIs describing multiple molecules
self.split_inchi_index = defaultdict(set)
for ci, inchi in tqdm(self.compound_inchis.items(),
leave=True,
position=0,
desc='Split InChIs'):
mol = rdkit.Chem.MolFromInchi(inchi)
if mol is None:
continue
try:
for m in rdkit.Chem.rdmolops.GetMolFrags(mol,
asMols=True): # split
i = rdkit.Chem.MolToInchi(m)
c = ic.inchi_conn_layer(i)
if not c in self.conn_split_inactive_inchis:
self.split_inchi_index[i].add(ci) # add to index
except:
pass
self.split_inchi_index = dict(self.split_inchi_index)
# extract connectivity layer of InChI and build index
self.inchi_connectivity_index = defaultdict(set)
for ci, inchi in tqdm(self.compound_inchis.items(),
leave=True,
position=0,
desc='Connectivity layer InChI index'):
try:
inchi = ic.strip_inchi(
inchi, exclude_inchis=self.conn_split_inactive_inchis
) # filter inchi
except:
pass
if inchi:
c = ic.inchi_conn_layer(inchi)
self.inchi_connectivity_index[c].add(ci)
self.inchi_connectivity_index = dict(self.inchi_connectivity_index)
self.inchi_split_connectivity_index = defaultdict(set)
for inchi, mols in tqdm(self.split_inchi_index.items(),
leave=True,
position=0,
desc='Split connectivity layer InChI index'):
c = ic.inchi_conn_layer(inchi)
# if not c in conn_split_inactive_inchis:
self.inchi_split_connectivity_index[c].update(mols)
self.inchi_split_connectivity_index = {
k: v
for k, v in self.inchi_split_connectivity_index.items()
if not k in self.conn_split_inactive_inchis
} # remove innactive InChIs
chembl_cursor.close()