Chemical Identifiers: CAS, CID, SID, SMILES, InChI and InChIKey

Chemical names are ambiguous in both directions: one substance carries a systematic name, several common names and often a research code, while similar names can belong to genuinely different substances. Identifiers exist to remove that ambiguity, but they do it in different ways and at different levels of precision. Knowing which level an identifier operates at is what stops two records being treated as the same record.

This is the distinction most often collapsed, and the one with the most practical consequence. PubChem holds two databases. Substance records hold chemical descriptions as individual depositors submitted them — versioned sample descriptions, kept without normalisation processing, essentially as provided and interpreted. Compound records hold unique chemical structures derived from those deposits by automated standardisation, which verifies that a structure is chemically sensible, recognises equivalent chemicals between depositors, and generates a preferred chemical representation. A SID is somebody’s submission; a CID is the normalised structure that submission was found to represent.

• SID identifies a deposit — one contributor’s description of a sample
• CID identifies a standardised structure derived from such deposits
• The mapping is many-to-one: many substance records can resolve to one compound record
• Because it is many-to-one, a count of substance records is not a count of independent sources
• Citing a CID cites a structure; citing an SID cites a particular deposit of it

The published description of PubChem’s standardisation reports that around 0.36% of submitted structures are rejected outright — overwhelmingly for atom valences that cannot be corrected without going back to the depositor — and that roughly 44% of those that pass are modified during the process. Tautomer canonicalisation and aromaticity standardisation account for most of the modification. The practical reading is that a depositor’s record and the compound record it maps to may not be identical, and that a vendor deposit is not an authoritative structure simply because it exists in the database.

A CAS Registry Number is assigned when a substance is registered, and is widely used in safety documentation, procurement and regulatory contexts. It is a registry key rather than a computed value, so it cannot be derived from a structure and carries no structural information in the digits themselves. Registration practice also means a salt, a hydrate and a free base are commonly separate registrations — so a CAS number can be more specific than the name beside it.

• Format is a sequence of digits in three hyphenated parts, with a final check digit
• Assigned on registration; not computable from a structure
• Salts, hydrates and stereoisomeric forms often carry their own numbers
• A number quoted against a name may describe a different form of that substance

These are notations rather than registries: each is computed from a structure, so any two people with the same structure can generate the same string without consulting a central authority. SMILES is compact and human-readable but admits many valid ways to write the same molecule, which is why canonical forms exist. InChI is designed to be a single standard string for a given structure, built in layers. An InChIKey is a fixed-length hash of an InChI, which makes it convenient to search on but means it cannot be read back into a structure.

A molecular formula counts atoms and a molecular weight sums their masses. Neither distinguishes isomers, and for peptides in particular the same formula covers every rearrangement of the same residues. They are useful checks — a formula inconsistent with a claimed identity is a real signal — but they cannot serve as identifiers, and a record matching only on formula has not been matched.

• Isomers share a formula and a weight
• Peptide sequence variants of identical composition are formula-identical
• Average and monoisotopic weights differ, and reports do not always say which is quoted
• A salt form changes the formula and the weight while the name may not change

Different resources apply different conventions to the same underlying substance: whether a salt or the free form is the primary record, how stereochemistry is represented when a source is ambiguous, which tautomer is treated as canonical, and how a name maps to a structure. Disagreement between two databases is usually a difference in convention or in record scope rather than an error, and it is resolved by comparing structures rather than names — which is what the notation identifiers are for. The mechanisms behind those differences, and how to tell a convention apart from a mistake, are set out in full in Why Scientific Databases Disagree.