def update_chebi_references():
# The reference table contains all the automated mappings from ChEBI
# IDs to IDs in other databases, except CAS, which only has manually
# curated mappings available in the database_accession table
# (see implementation in update_cas_to_chebi).
logger.info('--Updating ChEBI entries----')
url = 'ftp://ftp.ebi.ac.uk/pub/databases/chebi/' + \
'Flat_file_tab_delimited/reference.tsv.gz'
fname = os.path.join(path, 'reference.tsv.gz')
urlretrieve(url, fname)
with gzip.open(fname, 'rb') as fh:
logger.info('Loading %s' % fname)
df = pandas.read_csv(fh, sep='\t', index_col=None,
parse_dates=True, encoding='latin-1')
# Save PubChem mapping
fname = os.path.join(path, 'chebi_to_pubchem.tsv')
logger.info('Saving into %s' % fname)
df_pubchem = df[df['REFERENCE_DB_NAME']=='PubChem']
df_pubchem.sort_values(['COMPOUND_ID', 'REFERENCE_ID'], ascending=True,
inplace=True)
df_pubchem.to_csv(fname, sep='\t', columns=['COMPOUND_ID', 'REFERENCE_ID'],
header=['CHEBI', 'PUBCHEM'], index=False)
# Process PubChem mapping to eliminate SID rows and strip CID: prefix
# If the second column of the row starts with SID:, ignore the row
# If the second column of the row starts with CID:, strip out the CID
# prefix Otherwise, include the row unchanged
original_rows = read_unicode_csv(fname, '\t')
new_rows = []
for original_row in original_rows:
if original_row[1].startswith('CID:'):
new_row = original_row
new_row[1] = new_row[1][5:] # Strip out CID:
new_rows.append(new_row)
elif original_row[1].startswith('SID:'):
# Skip SID rows
continue
else:
# Include other rows unchanged
new_rows.append(original_row)
write_unicode_csv(fname, new_rows, '\t')
# In another round of cleanup, we try dealing with duplicate mappings in a
# principled way such that many-to-one mappings are allowed but one-to-many
# mappings are eliminated
original_rows = read_unicode_csv(fname, '\t')
chebi_pubchem = defaultdict(list)
pubchem_chebi = defaultdict(list)
for chebi_id, pc_id in original_rows:
chebi_pubchem[chebi_id].append(pc_id)
pubchem_chebi[pc_id].append(chebi_id)
# Looking for InChIKey matches for duplicates in the ChEBI -> PubChem
# direction
logger.info('Getting InChiKey matches for duplicates')
ik_matches = set()
for chebi_id, pc_ids in chebi_pubchem.items():
if len(pc_ids) > 1:
ck = chebi_client.get_inchi_key(chebi_id)
for pc_id in pc_ids:
pk = pubchem_client.get_inchi_key(pc_id)
if ck == pk:
ik_matches.add((chebi_id, pc_id))
# Looking for InChIKey matches for duplicates in the PubChem -> ChEBI
# direction
for pc_id, chebi_ids in pubchem_chebi.items():
if len(chebi_ids) > 1:
pk = pubchem_client.get_inchi_key(pc_id)
for chebi_id in chebi_ids:
ck = chebi_client.get_inchi_key(chebi_id)
if ck == pk:
ik_matches.add((chebi_id, pc_id))
rows = read_unicode_csv(fname, '\t')
header = next(rows)
header.append('IK_MATCH')
new_rows = [header]
for chebi_id, pc_id in rows:
if (chebi_id, pc_id) in ik_matches:
new_rows.append([chebi_id, pc_id, 'Y'])
else:
new_rows.append([chebi_id, pc_id, ''])
write_unicode_csv(fname, new_rows, '\t')
# Save ChEMBL mapping
fname = os.path.join(path, 'chebi_to_chembl.tsv')
logger.info('Saving into %s' % fname)
df_chembl = df[df['REFERENCE_DB_NAME']=='ChEMBL']
df_chembl.sort_values(['COMPOUND_ID', 'REFERENCE_ID'], ascending=True,
inplace=True)
df_chembl.to_csv(fname, sep='\t', columns=['COMPOUND_ID', 'REFERENCE_ID'],
header=['CHEBI', 'CHEMBL'], index=False)
def test_get_inchi_key():
ik = pubchem_client.get_inchi_key('5280613')
assert ik == 'JLVSPVFPBBFMBE-HXSWCURESA-O', ik
