Reproducible MS/MS library cleaning pipeline in matchms

Area of Science:

• Analytical Chemistry
• Computational Biology
• Bioinformatics

Background:

• Mass spectral libraries are crucial for mass spectrum annotation and training machine learning (ML) algorithms.
• Public mass spectrometry data libraries often lack sufficient metadata curation and harmonization, impacting ML model training.
• Variability in data quality poses challenges for developing robust ML models in mass spectrometry.

Purpose of the Study:

• To present a user-friendly, flexible, and reproducible pipeline for cleaning tandem mass spectrometry (MS/MS) library data.
• To enhance the quality of public mass spectral libraries for improved library searching and ML training datasets.
• To validate structure annotations and correct existing library data.

Main Methods:

• Development of a novel library cleaning pipeline specifically for tandem mass spectrometry data.
• Implementation of principles including ease of use, flexibility, and reproducibility.
• Incorporation of functionality for curating, correcting, and validating annotated libraries.

Main Results:

• The pipeline effectively cleans tandem mass spectrometry library data.
• Improved data quality in public mass spectral libraries.
• Enhanced accuracy and reliability of ML training datasets derived from these libraries.

Conclusions:

• The developed pipeline offers a high-quality solution for cleaning MS/MS libraries.
• It has the potential to become a standard tool in the field, improving reproducibility and data integrity.
• Cleaner libraries will advance mass spectrometry-based research, particularly in ML applications.