I’m seeing an increasing number of useful utilities being developed using AI coding tools
NISPO
NISPO is an open-source, RDKit-based Python package for generating IUPAC names of molecules. Generated names are designed to pass an OPSIN round-trip test, in which OPSIN parses the name back to the same molecular structure. The entire NISPO codebase was generated by OpenAI Codex using GPT-5.5.
In this case OPSIN provides the “scoring function” that allows the agent to optimise the structure to name function.
There are a couple of rule-based packages that undertake the structure to name function
Orthonym: verified IUPAC names for chemical structures in the wild.
Orthonym: a generator of verified IUPAC names for chemical structures. This open, deterministic, rule-based generator emits a name only after verifying that it correctly describes the input structure. OPSIN reads back each candidate name, and the InChIKey of the parsed structure must equal that of the input; otherwise, the engine declines to name.
openclature is similar
Morg2SMILES
In a similar manner Morg2SMILES recovers SMILES strings from Morgan fingerprints, in this case RDKit recomputes any candidate’s fingerprint and compares it bit-for-bit against the target. Claude wrote all the code.
The final model, trained on all of ChEMBL, scored on the common protocol every night used: 2,000 held-out molecules, 1,956 of them with fingerprints appearing nowhere in any training set, 20 guesses, sampling temperature T=0.6. Returns the original molecule, not merely one with a matching fingerprint: 96.6% within 20, 88.2% first guess
