CSMILES: A Compact, Human-Readable SMILES Extension for Conformations.

Area of Science:

• Computational Chemistry
• Cheminformatics
• Structural Biology

Background:

• Existing molecular line notation schemes like SMILES lack the ability to differentiate between various conformations of the same molecule.
• Representing and storing 3D molecular structures efficiently is crucial for computational chemistry and drug discovery.

Purpose of the Study:

• To develop an extension to the SMILES notation, termed CSMILES, capable of encoding specific molecular conformer information.
• To create a compact and canonical text-based representation for 3D molecular structures, including dihedral angles.

Main Methods:

• CSMILES extends the standard SMILES by incorporating dihedral angle values directly into the bond notation within the string.
• Canonicalization at the conformer level ensures invariance to atom reordering, translation, and rotation.
• A two-way conversion tool between 3D structures and CSMILES has been implemented, supported by Python code.

Main Results:

• CSMILES strings provide a compact and readable representation of molecular conformers, even for complex structures involving branching, rings, and symmetry.
• Canonical CSMILES strings are invariant to common transformations, ensuring consistent representation.
• A functional Python implementation facilitates conversion between 3D molecular data and CSMILES notation.

Conclusions:

• CSMILES offers a novel and efficient method for representing and storing 3D molecular structure information.
• The developed notation has potential applications in developing machine learning models for conformation-dependent molecular properties.