Skip to content
TheChatArchiveThe Chat Archive
Claudeclaude-opus-4-20250514· AnthropicHugging Face LMArena Human Preference 140K

Does SMILES and 2D Graph Representation have the same amount of information regarding a…

From a Chatbot Arena battleThe same person sent this prompt to two anonymous models at once. Compare it with gpt-4.1-mini-2025-04-14’s answer.
  1. User

    Does SMILES and 2D Graph Representation have the same amount of information regarding a molecule?

  2. Claude Assistant

    SMILES and 2D graph representations don't contain exactly the same amount of information about a molecule, though they're closely related.

    SMILES (Simplified Molecular Input Line Entry System) is a linear text notation that encodes molecular structure. A 2D graph representation shows atoms as nodes and bonds as edges, typically with spatial arrangement on a plane.

    Here are the key differences in information content:

    SMILES typically contains:

    • Atom types and connectivity
    • Bond types (single, double, triple, aromatic)
    • Branching patterns
    • Ring closures
    • Basic stereochemistry (using @, @@, /, \ notations)
    • Implicit hydrogen atoms (usually)

    2D graph representations typically contain:

    • All the connectivity information from SMILES
    • Explicit 2D coordinates for each atom
    • Visual bond angles and molecular shape
    • Sometimes explicit hydrogen atoms
    • Often clearer representation of aromatic systems

    The main difference is that 2D graphs include spatial layout information that SMILES lacks. However, this 2D arrangement is often algorithmically generated rather than reflecting true molecular geometry, so it's not necessarily "more accurate" information.

    Both representations can be converted to each other without loss of chemical identity, but you'd lose the 2D coordinates when going from graph to SMILES. Some subtle stereochemical information might also be represented differently or lost in conversion, depending on the specific implementation.

    For most cheminformatics applications, they're considered informationally equivalent regarding molecular identity, just with different encodings of the same underlying chemical structure.

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.