LAS: A Lipid Annotation Service Capable of Explaining the Annotations It Generates

Area of Science:

• Biochemistry
• Metabolomics
• Bioinformatics

Background:

• Lipid annotation is crucial for understanding biological processes.
• Existing tools often lack transparency in their annotation processes.
• Researchers require reliable methods for validating lipid identities.

Purpose of the Study:

• To introduce the Lipid Annotation Service (LAS), a REST API for lipid annotation.
• To enhance the reliability and interpretability of lipid annotation results.
• To provide automated evidence-based rationale for annotation certainty levels.

Main Methods:

• Development of a REST API service (LAS) incorporating rules from CEMBIO and literature.
• Implementation of an automated evidence extraction mechanism based on inter-rule relationships.
• Assignment of certainty levels (very unlikely, unlikely, likely, very likely) to annotations.
• Natural language explanation of supporting or refuting evidence for annotations.

Main Results:

• LAS is the first metabolite annotation tool to provide natural language explanations for its annotations.
• The service automatically extracts and presents evidence to support or refute annotations.
• Evaluation on blood plasma samples demonstrated an F-measure exceeding 80% for LAS.
• Increased user understanding and confidence in lipid annotation results.

Conclusions:

• LAS significantly improves the process of lipid annotation by providing transparent, evidence-based certainty levels.
• The tool enhances researcher confidence and understanding through natural language explanations.
• LAS demonstrates high performance and reliability in lipid identification tasks.