An Automated Model Annotation System (AMAS) for SBML Models.

Area of Science:

• Biochemistry
• Systems Biology
• Computational Biology

Background:

• Biochemical models require detailed annotations for chemical species and reactions to ensure clarity and define limitations.
• Current biochemical models often lack sufficient annotations, hindering their understanding and application.
• Existing gene annotation tools are not directly applicable to biochemical models due to technical differences.

Approach:

• Developed the Automated Model Annotation System (AMAS) to predict annotations for Systems Biology Markup Language (SBML) model elements.
• Implemented a general framework using annotated reference databases and a similarity-based match score function.
• Instantiated the framework for species and reactions using databases like ChEBI and BiGG, with string similarity for matching.

Key Points:

• AMAS demonstrates sub-second response times for annotation prediction.
• Prediction accuracy ranges from 80% to 95% for species and reactions in BiGG and BioModels.
• AMAS is available as an open-source, pip-installable Python package and command-line tool.

Conclusions:

• AMAS effectively improves the quality and quantity of annotations in biochemical models.
• The system offers a practical solution for annotating species and reactions within SBML models.
• AMAS enhances the usability and interpretability of biochemical models in systems biology research.