MSCAT: A Machine Learning Assisted Catalog of Metabolomics Software Tools

Area of Science:

• Computational Biology
• Bioinformatics
• Data Science

Background:

• Metabolomics data analysis is hindered by the lack of accessible and usable software tools.
• Hundreds of diverse, open-source metabolomics tools exist, posing challenges for researchers in selection and adoption.
• There is a growing need for guidance on choosing analytical tools and ensuring reproducibility in metabolomics.

Purpose of the Study:

• To develop a sustainable and continuously updated database of metabolomics software tools.
• To assist researchers in selecting appropriate data analysis workflows for their specific metabolomics studies.
• To semi-automate the identification and cataloging of new metabolomics software tools from scientific literature.

Main Methods:

• Developed the Metabolomics Software CATalog (MSCAT) database.
• Implemented a machine learning (ML) approach using Named Entity Recognition (NER) with a neural network (CNN-LSTM-CRF) to identify software tool names in abstracts.
• Designed an end-user interface for filtering tools and visualizing the metabolomics software landscape.

Main Results:

• Created a comprehensive catalog of metabolomics software tools.
• Successfully semi-automated the discovery of new tools, reducing manual curation effort.
• Provided a searchable and filterable database to aid researchers in tool selection.

Conclusions:

• MSCAT addresses the bottleneck in metabolomics data analysis by providing a centralized resource for software discovery.
• The ML-driven approach enhances the efficiency and scalability of maintaining an up-to-date catalog.
• MSCAT facilitates informed decision-making for metabolomics researchers, promoting best practices and reproducibility.