STOUT V2.0: SMILES to IUPAC name conversion using transformer models.

Area of Science:

• Computational Chemistry
• Cheminformatics
• Artificial Intelligence in Chemistry

Background:

• Systematic chemical naming follows International Union of Pure and Applied Chemistry (IUPAC) rules, which are complex and challenging to apply consistently.
• Accurate translation from Simplified Molecular Input Line Entry System (SMILES) strings to IUPAC names is vital for streamlining chemical structure naming.

Purpose of the Study:

• To present STOUT (SMILES-TO-IUPAC-name translator) V2, a novel transformer-based model for translating SMILES notation to IUPAC names.
• To demonstrate the potential of neural networks in chemical nomenclature, complementing existing deterministic algorithms.

Main Methods:

• Development of a transformer-based neural network model (STOUT V2).
• Training the model on a large dataset of approximately 1 billion SMILES strings and their corresponding IUPAC names.
• Leveraging OpenEye's Lexichem software for access to chemical naming functionalities.

Main Results:

• STOUT V2 achieves exceptional accuracy in generating IUPAC names, even for complex chemical structures.
• The model effectively captures intricate patterns and relationships within chemical structures for precise naming.
• A web application was developed to enhance accessibility, and the model/source code are open-source.

Conclusions:

• Transformer-based neural approaches, like STOUT V2, show significant potential for chemical nomenclature.
• STOUT V2 offers a powerful, accurate, and accessible tool to aid chemists in systematic chemical naming.
• The open-source nature of STOUT V2 encourages further research and development in computational chemistry.