Canonical smiles generation

[sara-hashemi]

I am using the rdkit's Chem library and importing AllChem to generate canonical smiles from inchi, InChIKey and PC_CID. When comparing the generated canonical smiles based on inchi and PubChem CID, the smiles don't match. I would like to see why the smiles generated through inchi do not match with those generated by pubchem CID or InChIKey.

[jasondbiggs]

How do you make a SMILES string from an InChIKey or CID number?

[sara-hashemi]

Both are based on pubchempy. I use the below function to get the smiles string from CID:

import pubchempy as pcp
def get_smiles_from_cid(cid):
try:
compound = pcp.Compound.from_cid(cid)
return compound.canonical_smiles if compound else None
except Exception as e:
print(f"Error fetching PubChem SMILES for CID {cid}: {e}")
return None

And the below code to generate the canonical smiles from InChIKey:

import pubchempy as pcp
def get_smiles_from_inchikey(inchikey):
try:
compound = pcp.get_compounds(inchikey, 'inchikey')
if compound:
return compound[0].canonical_smiles
except Exception as e:
print(f"Error fetching PubChem SMILES for InChIKey {inchikey}: {e}")
return None

[sara-hashemi]

I use the below function to get the smiles string from CID:

import pubchempy as pcp
def get_smiles_from_cid(cid):
try:
compound = pcp.Compound.from_cid(cid)
return compound.canonical_smiles if compound else None
except Exception as e:
print(f"Error fetching PubChem SMILES for CID {cid}: {e}")
return None

And the below code to generate the canonical smiles from InChIKey:

import pubchempy as pcp
def get_smiles_from_inchikey(inchikey):
try:
compound = pcp.get_compounds(inchikey, 'inchikey')
if compound:
return compound[0].canonical_smiles
except Exception as e:
print(f"Error fetching PubChem SMILES for InChIKey {inchikey}: {e}")
return None

[bp-kelley]

The short answer is that pubchem's canonical smiles were created via the OpenEye toolkit ( a closed source commercial toolkit ) and there is no expectation that different canonicalization algorithms generate the same smiles. There is no "one standard" canonicalization, each toolkit has a different implementation.

In general, if you are using your own database or records, you need to choose a canonicalization that you can reproduce.

>>> import pubchempy as pcp
>>> from rdkit import Chem
>>> cid = 5381226
>>> compound = pcp.Compound.from_cid(cid)
>>> compound.isomeric_smiles
'C[C@H]1/C=C/C=C(\\C(=O)NC\\2=C(C3=C(C4=C(C(=C3O)C)O[C@@](C4=O)(O/C=C/[C@@H]([C@H]([C@H]([C@@H]([C@@H]([C@@H]([C@H]1O)C)O)C)OC(=O)C)C)OC)C)C(=O)/C2=C/NN5CCN(CC5)C)O)/C'
>>> Chem.CanonSmiles(compound.isomeric_smiles) == compound.isomeric_smiles
False