STOUT: SMILES to IUPAC names using neural machine translation

Area of Science:

• Computational chemistry
• Cheminformatics
• Artificial intelligence in chemistry

Background:

• Chemical compounds require standardized identification methods, such as graphical depictions, string representations, or chemical names.
• The International Union of Pure and Applied Chemistry (IUPAC) established a naming scheme, but its complexity makes manual assignment difficult.
• Existing automated tools for chemical name assignment are limited.

Purpose of the Study:

• To present STOUT (SMILES-TO-IUPAC-name translator), a novel deep-learning approach for chemical name translation.
• To enable automated generation of IUPAC names from SMILES strings.
• To facilitate the reverse translation, predicting SMILES strings from IUPAC names.

Main Methods:

• Utilized a deep-learning neural machine translation model.
• Trained the model for bi-directional translation between SMILES strings and IUPAC names.
• Evaluated performance using BLEU score and Tanimoto similarity index.

Main Results:

• Achieved an average BLEU score of approximately 90% for both translation directions.
• Demonstrated a Tanimoto similarity index exceeding 0.9 for predicted compounds.
• Observed high similarity even in incorrect predictions, indicating model robustness.

Conclusions:

• STOUT offers an effective deep-learning solution for automated chemical name translation.
• The model accurately converts between SMILES and IUPAC nomenclature.
• The high performance suggests STOUT's potential to aid cheminformatics tasks and chemical data management.