Can not generate SMILES of compounds from known CAS numbers

[MathewGolding]

Hello, I recently began using PubChemPy, and thank you for creating this incredible platform! I am attempting to find the SMILES of over 500 compounds via the get_compound() function however the only data for my input into the function is a I have is a list of CAS numbers. As far as I can tell, PubChemPy has no way of pulling a compound from the CAS number using the get_compound() function as CAS numbers are not a valid input for this function. i.e. I have the CAS number "3586-12-7" yet there is not a valid input for this value into the get_compound() function

Any help resolving this is massively appreciated!

[nbehrnd]

In a virtual environment of Python, amended by cirpy (version 1.0.2 from pypi), I just run the following conversions with the CAS of the example here; your compound; glucose and fructose as reported by Wikipedia's property boxes. It could be a suitable tool to extend from here:

$ python
Python 3.11.7 (main, Dec  8 2023, 14:22:46) [GCC 13.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import cirpy
>>> 
>>> data = ['108-95-2', '3586-12-7', '50-99-7', '57-48-7']
>>> 
>>> for entry in data:
...     cirpy.resolve(entry, 'smiles')
... 
'Oc1ccccc1'
'Nc1cccc(Oc2ccccc2)c1'
'OCC1OC(O)C(O)C(O)C1O'
'OC[C@@H](O)[C@@H](O)[C@H](O)C(=O)CO'
>>> 

[MathewGolding]

Tried out CIRpy as recommended by nbehrnd above. Whilst it was, for the most part, successful I still lack the SMILES of roughly 70 compounds. Is there anything else I can use, preferably with PubChemPy to get these SMILES?

[nbehrnd]

Well, it depends a bit on the chemicals behind the CAS number.

• some chemical suppliers equally provide a SMILES string, buried in the description of properties. This is practice is not consistent among all suppliers, the algorithm to assign a SMILES string can be a different one each time. See e.g., Sigma about one form of glucose:

  

• the English edition of Wikipedia describes chemicals with their CAS number (property pox). If you have the chemical name of the structures (or the [sub]structure), you can tap into a set of about 22k chemicals (if I remember correctly); the Wikipedia Chemical Structure explorer can be of help here.

• around/after Wikipedia and ACS/The American Chemical Society settled an agreement to use the CAS registry numbers in Wikipedia, ACS started a project "common chemistry". The summaries' web addresses follow a pattern like https://commonchemistry.cas.org/detail?cas_rn=50-99-7. Speculation: programmatic consultation and harvest of the SMILES from there could be automated (Python's beautifulsoup, or an other package of your preference). Coverage by the number of CAS entries is considerably larger (once advertised to include approx. 500k records) and still free.

• chemspider by the Royal Society of Chemistry can be queried e.g., by a pattern like http://www.chemspider.com/Search.aspx?q=50-99-7, or with an API proper (link). As you see for the results on the example of glucose, there may be multiple structures per CAS (well, ...) but from a results page like this, below the icon of the floppy disc is a .mol file which can be processed further (vide infra). The database claims to store 100 million records.

---

• If you work for a university, or/and have access to a research library with a paid subscription to databases like chemical abstracts/SciFinder scholar (by ACS), or Reaxys (provided by Elsevier), you can perform an individual search by CAS number and request the output of a .mol/.sdf file. An application like RDKit, or openbabel can then convert (batches of) them into one form of SMILES.

[khoivan88]

@MathewGolding , I made something that lookup pubchem cid based on different identifier (CAS, smiles, inchi, inchi_key). Maybe you can draw to your application: https://github.com/khoivan88/pka_lookup/blob/9705117d70e9162fe4c410d5fa884d550f02bad2/src/pka_lookup_pubchem.py#L44

README: https://github.com/khoivan88/pka_lookup/tree/master