Banner

IMPORTANT NEWS

We're thrilled to share the release of our new preprint introducing MS2LDA 2.0: "Large-scale discovery and annotation of hidden substructure patterns in mass spectrometry profiles". The new MS2LDA 2.0 is available at: https://github.com/vdhooftcompmet/MS2LDA. The results can be visualized at: https://ms2lda.org/viz/ or via the button above.

What's new in MS2LDA 2.0?

  • ⚡ Up to 14x faster unsupervised Mass2Motif discovery—scalable to large datasets
  • 🧠 Mass2Motif Annotation Guidance (MAG) for high-accuracy structural annotation
  • 📊 A modern, Dash-based interface for exploring Mass2Motifs: richer plots and smoother navigation
  • 🔍 Seamless integration with MassQL-searchable MotifDB

Together, these enhancements empower deeper exploration of hidden chemical patterns in MS data.

Example datasets and how to run the tool:

We applied MS2LDA 2.0 to:

  • A tomato-spiked metabolomics dataset
  • A medicinal fungal dataset

Both are available at: 📁 https://zenodo.org/records/15688609

To reproduce our analyses and explore further, see our example notebooks: 📓 https://github.com/vdhooftcompmet/MS2LDA/tree/main/notebooks

For MAG, we retrained Spec2Vec (positive mode) using updated data available at: 📁 https://zenodo.org/records/12543129

⚠️ Notice: The legacy MS2LDAviz application will be retired soon. Please update your workflows and bookmarks to the new system.

Thank you to our amazing community for your ongoing support, and to the University of Glasgow MLVS Shared Research Facility (formerly Glasgow Polyomics) for their generous contributions to this project.

11th October 2023: Thanks to the continuous support from Glasgow Polyomics, we've recently migrated MS2LDA to a better virtual machine. This should hopefully allow for a more robust service in the future. Unfortunately during the migration process, all jobs submitted on the system from 25 September to now (11th October) were not working. If you tried to submit a job during that time period, your job will silently fail and/or it will show as “Pending” status forever. In that case, please delete that submitted analysis and try again as it should work fine now. If you still encounter a problem, please contact our team.

We also run on limited disk space and database capacity, so if you have any unused analyses please delete them. Doing this helps us sustain this free service for the metabolomics community.

15th June 2023: We've recently increased the disk storage of our server to address Internal Error 500 incidents that were caused by insufficient storage. However we would like to request our users to please delete old analyses that you no longer need. This will help make the server run smoothly for everyone.

13th July 2021: Ms2lda.org has now been running for nearly 4 years! 🎉 Thank you for your continuing support. MS2LDA is also integrated with GNPS, allowing you to map Mass2Motif occurrences in Molecular Families. Alternatively LDA analysis can be run on this website directly (subject to computational limits).

15th January 2021: Ms2lda.org is finally migrated to Python 3, and job submission is now working again! If you have any old, unused experiments, please delete them to help us save space.

25th November 2018: We have introduced many new features to make interpreting Mass2Motifs in MS2LDA.org easier:

  • A new upload function in Create Experiment lets you upload large locally-run LDA experiments — script under development.
  • MotifDB (a database of curated Mass2Motifs) has been incorporated, plus automated ClassyFire predictions using a neural network.
  • Automatic MAGMa annotations of fragment and neutral-loss features are now integrated.

4th August 2017: Our paper on MS2LDA variability across large experiments is published in Analytical Chemistry. We also introduced dual thresholds on probability and overlap_score for spectrum–motif links.

26th September 2017: Our application note is published in Bioinformatics. The original idea is detailed in PNAS 2016 and in Joe Wandy’s PhD thesis.