Skip to content
TheChatArchiveThe Chat Archive
gpt-4.1-mini-2025-04-14· OpenAIHugging 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 Claudeclaude-opus-4-20250514’s answer.
  1. User

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

  2. Assistant

    Great question! Let's break down the information content of SMILES and 2D graph representations of molecules:

    SMILES (Simplified Molecular Input Line Entry System)

    • What it is: A linear string notation that encodes the molecular structure.
    • Information content:
      • Encodes the connectivity of atoms (which atoms are bonded to which).
      • Includes bond types (single, double, triple, aromatic).
      • Can encode stereochemistry (chiral centers, cis/trans isomerism) if specified.
    • Limitations:
      • It is a linearized representation, so the spatial arrangement is implicit, not explicit.
      • Different SMILES strings can represent the same molecule (canonical SMILES tries to standardize this).
      • Does not explicitly encode 3D geometry or conformations.

    2D Graph Representation

    • What it is: A graph where nodes represent atoms and edges represent bonds.
    • Information content:
      • Explicitly shows connectivity and bond types.
      • Can include stereochemical annotations on nodes/edges.
      • Represents the molecule as a graph, which is a more natural representation of connectivity.
    • Limitations:
      • Usually does not encode 3D coordinates or conformations.
      • The "2D" layout is often just a planar drawing for visualization and does not represent actual spatial geometry.

    Comparison: Do they have the same amount of information?

    • Connectivity and bond types: Both encode this fully.
    • Stereochemistry: Both can encode stereochemistry, but it depends on the specific notation or graph annotations.
    • 3D structure: Neither inherently encodes 3D geometry or conformations.
    • Redundancy and uniqueness:
      • SMILES is a linear string and can have multiple valid forms for the same molecule unless canonicalized.
      • Graphs are more canonical in terms of connectivity but can vary in layout.

    Summary:

    • In terms of molecular connectivity and bond types, SMILES and 2D graph representations contain essentially the same information.
    • Neither inherently contains 3D spatial information.
    • Stereochemistry can be encoded in both but requires explicit notation.
    • SMILES is a linear encoding of the graph, so it is a different format but not fundamentally more or less informative about connectivity.

    If you want to capture 3D structure or conformations, you need additional data (e.g., 3D coordinates, molecular dynamics data), which neither SMILES nor simple 2D graphs provide.


    In short:
    SMILES and 2D graph representations are equivalent in the amount of structural information they encode about a molecule’s connectivity and bonding, but neither fully captures 3D geometry.

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