International Chemical Identifier

International Chemical Identifier (InChI) is an open-source, non-proprietary, canonical textual identifier for chemical substances designed by the International Union of Pure and Applied Chemistry (IUPAC) and the U.S. National Institute of Standards and Technology (NIST) to provide every molecular structure with a single, globally invariant digital signature across chemoinformatics repositories, scientific journals, and patent databases. Whereas a molecule can be represented by multiple valid Simplified Molecular-Input Line-Entry System (SMILES) strings depending on the software vendor's aromaticity and graph-traversal heuristics, the reference InChI algorithm normalizes valence states, tautomerism, and canonical atom numbering so that any chemical drawing or connection table produces the identical Standard InChI string (InChI=1S/...) regardless of the originating software.

Developed between 2000 and 2005 by Stephen R. Heller, Alan D. McNaught, Igor V. Pletnev, Stephen E. Stein, and Dmitrii V. Tchekhovskoi and governed by the non-profit InChI Trust, the identifier was augmented in 2007 (InChI version 1.02) with the InChIKey—a fixed-length, 27-character uppercase string derived via SHA-256 cryptographic hashing (AAAAAAAAAAAAAA-BBBBBBBBFV-P). Together, InChI and InChIKey serve as the primary relational link connecting PubChem and ChEMBL, the Protein Data Bank (PDB), UniChem, and CrossRef.

Bond-Order Elimination, Tautomer Normalization, and McKay's nauty Canonicalization

To eliminate representation ambiguities that cause traditional chemical registries to split a single compound across multiple records, the InChI generation pipeline executes three deterministic transformations on an input molecular graph G = (V, E):

1. Removal of Formal Bond Orders and Electrons: All localized single, double, triple, and aromatic bond designations are stripped, leaving only a skeletal sigma-bond connectivity graph of non-hydrogen (heavy) atoms together with the total count and distribution of hydrogen atoms and net electric charge. Because resonance hybrids (such as the two Kekulé structures of benzene, C₆H₆) share identical sigma connectivity and hydrogen counts, InChI bypasses vendor-dependent Hückel aromaticity perception.
2. Mobile-Hydrogen (Tautomeric) Group Normalization: In Standard InChI, hydrogen atoms capable of migrating rapidly between heteroatoms in solution—such as keto–enol, amide–imidic acid, amidine, or purine/pyrimidine lactam–lactim tautomers—are pooled into shared mobile-hydrogen groups ((H,num1,num2,...)). Consequently, 2-hydroxypyridine and pyridin-2(1*H*)-one map to the identical Standard InChI.
3. Canonical Graph Labeling via nauty: Canonical atom numbers (1, 2, ..., N) are assigned to heavy atoms by solving the graph automorphism and canonical labeling problem using Brendan D. McKay's nauty ("No Automorphisms, Yes?") algorithm. Atoms are partitioned by element in Hill-system order and refined iteratively until a lexicographically minimal adjacency table is obtained.

The Six Structural Layers of an InChI String

Every InChI string begins with a version prefix (InChI=1S/ for Standard InChI or InChI=1/ for non-standard InChI) followed by up to six hierarchical layers separated by forward slashes (/). Within each layer, sublayers are prefixed by a single lowercase character code (c, h, q, p, b, t, m, s, i, f, r):

1. Main Layer (Mandatory): Always present. Starts with the Hill-order chemical formula (carbon first, hydrogen second, then all remaining elements alphabetically, e.g., C8H10N4O2 for caffeine), followed by the /c skeletal connectivity sublayer (listing canonical atom adjacency chains and branches without bond orders) and the /h hydrogen sublayer (specifying how many immobile hydrogens are attached to each canonical heavy atom and listing any mobile tautomeric hydrogen pools).
2. Charge Layer (/q and /p): Records the net formal charge of each component (/q+1, /q-1) and the number of protons (H+) added or removed (/p+1, /p-1) during acid–base normalization.
3. Stereochemical Layer (/b, /t, /m, /s): Encodes 3D configuration independently of atom-numbering conventions: /b specifies double-bond and cumulene E/Z parity (+ or - based on canonical neighbor ranks); /t specifies tetrahedral sp³ parity (+, -, or ?) at each chiral center; /m (0 or 1) indicates whether the absolute enantiomeric configuration was inverted during canonicalization; and /s (1 = absolute, 2 = relative/racemic, 3 = meso) flags the stereochemical type.
4. Isotopic Layer (/i): Lists isotopic mass deviations from rounded standard atomic weights (e.g., /i1+1 for ¹³C at canonical atom 1 or /i/hD for deuterium substitution) along with isotopic stereocenters.
5. Fixed-H Layer (/f, Non-Standard Only): Omitted in Standard InChI (1S), this optional layer pins mobile hydrogens to specific heteroatoms when an application requires distinguishing individual keto vs. enol or guanine vs. isoguanine tautomers.
6. Reconnected Layer (/r, Non-Standard Only): By default, Standard InChI disconnects bonds between metal atoms and ligands in coordination complexes; when the /RecMet flag is invoked, the /r layer restores explicit organometallic coordination bonds.

Anatomy of the SHA-256 InChIKey

Because a full InChI string for a complex natural product or polypeptide can exceed 300 to 1,000 characters and contains punctuation (/, (, ), ,, -, +) that complicates URL routing and search-engine tokenization, IUPAC and the InChI Trust introduced the InChIKey in 2007. Every InChIKey has a fixed 27-character format (AAAAAAAAAAAAAA-BBBBBBBBFV-P) divided into three hyphen-separated blocks of uppercase ASCII letters:

• First Block (14 characters, AAAAAAAAAAAAAA): A base-26 encoding of the truncated SHA-256 hash of the Main connectivity layer (formula, /c, and /h sublayers). Any two stereoisomers, isotopologues, or protonation states that share the same 2D covalent skeleton and tautomer group share an identical 14-character first block. For example, both L-alanine (QNAYBMKLOCPYGJ-REOHCLBHSA-N) and D-alanine (QNAYBMKLOCPYGJ-UWTATZPHSA-N) begin with QNAYBMKLOCPYGJ.
• Second Block (10 characters, BBBBBBBBFV): The first 8 characters (BBBBBBBB) are a truncated SHA-256 hash of all remaining layers (stereochemistry, isotopic substitution, charge, and non-standard fixed-H/metal layers; UHFFFAOY indicates an achiral, non-isotopic structure). Character 9 (F) is the Standardization Flag (S for Standard InChI 1S, N for non-standard InChI). Character 10 (V) is the Version Character (A for InChI version 1).
• Third Block (1 character, P): Encodes the net protonation state of the parent molecule from the /p sublayer (N = neutral/no proton adjustment; O, P, Q... = +1, +2, +3 protons added; M, L, K... = −1, −2, −3 protons removed).

Extensions: RInChI for Reactions and Polymer/Mixture InChIs

To extend canonical identifiers beyond discrete small molecules, the InChI Trust released the Reaction InChI (RInChI) standard (Grethe et al., 2018), which canonically orders reactants, products, catalysts, and solvents alongside a directional reaction-hash key (RInChIKey), as well as experimental MInChI (mixtures) and polymeric InChI notations for macromolecular repeat units.

How does InChI differ from SMILES, and why are both used in chemoinformatics?

SMILES and InChI solve complementary problems in chemoinformatics. SMILES is designed for human readability, compact storage, fast depth-first graph reconstruction, and regular-expression substructure querying (via SMARTS), making it the preferred representation for chemical drawing programs, reaction transforms, and generative neural networks. However, SMILES canonicalization varies across software vendors. InChI is not meant to be written by hand; instead, it strips localized bond orders and normalizes mobile hydrogens via McKay's nauty graph isomorphism algorithm so that every database worldwide generates the exact same identifier (InChI and InChIKey) for a given chemical substance.

Can a chemical structure be reconstructed directly from an InChI or an InChIKey?

A complete InChI string (InChI=1S/...) is a reversible, lossless mathematical encoding of the normalized molecular graph: the official IUPAC InChI-to-Structure utility can deterministically reconstruct the full heavy-atom connectivity, hydrogen placement, formal charges, and 3D stereocenters from an InChI string alone. In contrast, an InChIKey (AAAAAAAAAAAAAA-BBBBBBBBFV-P) is a one-way cryptographic hash generated by truncating SHA-256 digests of the InChI layers; just as a file cannot be mathematically inverted from its SHA-256 checksum, converting an InChIKey back into a chemical structure requires querying a lookup resolver table such as the NCI/CADD Chemical Identifier Resolver, UniChem, or PubChem.

Why do some tautomers have the same Standard InChI, and how can specific tautomers be distinguished?

In aqueous solution at room temperature, prototropic tautomers—such as the keto and enol forms of 1,3-dicarbonyls, or the 1*H* and 3*H* imidazole ring tautomers of histidine—interconvert rapidly via proton transfer. Standard InChI (InChI=1S/...) intentionally groups mobile hydrogens shared across heteroatom donor–acceptor paths into a single /h(H,...) pool so that databases do not fragment bioassay data for the same equilibrating substance across separate entries. When a study requires distinguishing individual tautomers (for example, in gas-phase spectroscopy or molecular docking), generating a non-standard InChI with the -FixedH option appends the /f layer, producing a distinct InChI and an InChIKey flagged with N instead of S at position 24.

1. Heller, S. R., McNaught, A., Pletnev, I., Stein, S., & Tchekhovskoi, D. (2015). InChI, the IUPAC International Chemical Identifier. Journal of Cheminformatics, 7(1), 23. https://doi.org/10.1186/s13321-015-0068-4
2. Pletnev, I., Erin, A., McNaught, A., Blinov, K., Tchekhovskoi, D., & Heller, S. (2012). InChIKey collision resistance: An experimental testing. Journal of Cheminformatics, 4(1), 39. https://doi.org/10.1186/1758-2946-4-39
3. McKay, B. D., & Piperno, A. (2014). Practical graph isomorphism, II. Journal of Symbolic Computation, 60, 94–112. https://doi.org/10.1016/j.jsc.2013.09.003
4. Grethe, G., Blanke, G., Kraut, H., & Goodman, J. M. (2018). International chemical identifier for reactions (RInChI). Journal of Cheminformatics, 10(1), 22. https://doi.org/10.1186/s13321-018-0277-8