BigSMILES

Line notation for describing macromolecules

BigSMILES is a structurally based line notation for describing macromolecules, extending the small-molecule SMILES string to the stochastic domain of polymers.

History

It was introduced in 2019 by Tzyy-Shyang Lin and Bradley Olsen at MIT, with collaborators at Duke and Northwestern, and published in ACS Central Science.

Compared to SMILES

Where SMILES encodes a single deterministic molecular graph, real polymers are ensembles: a sample contains chains that differ in length, sequence, and connectivity. BigSMILES was designed to capture this uncertainty compactly so that polymers can be indexed, searched, and aggregated in databases and used as features in polymer informatics and machine learning.

Because it composes with ordinary SMILES, BigSMILES can describe homopolymers, random, block, and alternating copolymers, branched, hyperbranched, and graft architectures, ladder polymers via nested descriptors, rings, and networks.

Stochastic Object

The central construct is the stochastic object, a polymeric segment written inside curly braces that lists one or more repeat units separated by commas, optionally followed by a semicolon and a set of end groups. Rather than fixing chain length or monomer order, a stochastic object denotes the full set of molecular fragments consistent with the stated connection rules.

Syntax

Connectivity is expressed through bonding descriptors attached to boundary atoms of each repeat unit. AA-type descriptors, written [$] (optionally with a numeric index like [$1]), may join any other [$] of the same index, which suits vinyl and addition polymers; the conjugate AB-type descriptors [<] and [>] join only their opposite of matching index, which suits condensation and directional linkages. Terminal descriptors at the braces specify how the polymeric segment attaches to surrounding non-polymeric SMILES text or to explicit end groups.

Related Formats

Similar formats are InChI, SMARTS, and structure containers like Molfile and SDF. It is supported by the CRIPT polymer data ecosystem and tooling built on top of SMILES parsers like RDKit and Open Babel.

Limitations

• BigSMILES encodes topology and connection rules but not quantitative distributions, so molecular weight, dispersity, composition ratios, and tacticity fractions are not part of the base string and must be carried as external metadata.
• Canonicalization is also nontrivial, and later work has addressed canonical forms for defined-backbone polymers, coarse-grained models (CG-BigSMILES), and supramolecular non-covalent bonds.

It is thus best understood as a qualitative, ensemble-level structural identifier rather than a full physical specification of a polymer sample.