Good practices and recommendations for using and benchmarking computational metabolomics metabolite annotation tools

Area of Science:

• Metabolomics
• Computational Biology
• Mass Spectrometry

Background:

• Untargeted metabolomics using mass spectrometry provides comprehensive biological sample profiles but suffers from low metabolite annotation rates (around 10%).
• Low annotation rates hinder biochemical interpretation and comparative analysis of metabolomics studies.
• De novo structural characterization of mass spectral data is complex and time-consuming.

Purpose of the Study:

• To review recent advances in computational metabolite annotation workflows.
• To focus on the evaluation and comparison of different annotation tools.
• To provide recommendations for benchmarking and comparing novel tools.

Main Methods:

• Review of mass spectral library-based annotation methods.
• Discussion of machine learning-supported metabolite annotation workflows.
• Analysis of molecular networking and mass spectral similarity scores.

Main Results:

• Computational metabolomics, including molecular networking and machine learning, shows great promise for large-scale and reliable metabolite annotation.
• Significant progress has been made in computational metabolite annotation tools.
• Inconsistencies in benchmarking hinder the selection of optimal tools.

Conclusions:

• Recent developments in computational metabolomics are fundamentally changing workflows.
• Overcoming method performance ambiguities and annotation bottlenecks is achievable.
• Improved tools will enhance biochemical interpretation and enable effective comparison of metabolomics studies.