InChI

InChI (IUPAC International Chemical Identifier) is a non-proprietary textual identifier for chemical substances, designed so that a given molecular structure always maps to exactly one canonical string. It was developed jointly by IUPAC and the US National Institute of Standards and Technology (NIST) between roughly 2000 and 2005, and since 2009 has been stewarded by the InChI Trust, a UK-registered charity that maintains the reference generator software (an open C library) together with IUPAC working groups. The primary goal is unambiguous, machine-derivable identification: unlike a CAS registry number, an InChI is computed from the structure itself rather than assigned by an authority.

Structurally, an InChI string is a layered serialization prefixed with a version tag (e.g. "InChI=1S/", where "S" marks a Standard InChI). Successive layers, delimited by slashes, encode the molecular formula, the atom-connectivity table, hydrogen positions, and then optional charge, stereochemical, isotopic, fixed-hydrogen (tautomer) and reconnected-metal layers. Generation is a deterministic three-stage pipeline: normalization of the input, canonical numbering of atoms (a graph-canonicalization step that resolves the automorphism of the molecular graph), and serialization. Because the connectivity layer is effectively a canonicalized labeled graph, InChI is a genuine graph-structured chemical representation rather than a coordinate-based one.

For indexing at web and database scale, the companion InChIKey condenses the full string into a fixed 27-character token: a 14-character block hashing the molecular skeleton, a dash, an 8-character block hashing stereo/isotopic/tautomeric detail, and flag characters, all derived by truncating an SHA-256 hash. Collisions are possible in principle but empirically extremely rare, making the key a practical primary key for deduplication and cross-database linking.

In the format-conversion world, InChI is typically produced from connection-table or line-notation inputs such as format:molfile, format:sdf, and format:smiles, and it coexists with format:inchikey as its hashed lookup form. Toolkits including tool:rdkit and tool:openbabel generate and consume it, alongside the reference InChI library. Its strengths are canonical uniqueness, openness, and reliable structure-based deduplication across resources like PubChem and ChemSpider. Its limitations are equally concrete: it discards 3D coordinates and 2D depiction (so it cannot round-trip drawings), handles some organometallics, polymers, and mixtures poorly, and is only semi-human-readable, which is why editable formats like SMILES remain preferred for authoring and query construction.