Overview
DreaMS generates learned vector representations of tandem mass spectra for similarity analysis and downstream prediction. The project provides pretrained models, tutorials and a library-matching web demonstration; local embeddings can be reused in clustering and molecular-network analysis.
What can this tool do?
Common uses
- Generate learned embeddings for MS/MS spectra.
- Compare spectra using embedding similarity or build large-scale molecular networks.
- Fine-tune pretrained representations for task-specific prediction.
Scientific fit
Primary function
Spectral processing and comparison
Other functions
Machine learning and prediction, Molecular networking and chemical similarity, Metabolite annotation and identification
Analysis types
Untargeted metabolomics
MS level
MS/MS
Acquisition
Data-dependent acquisition (DDA), Targeted MS/MS, General or unspecified acquisition
Ion mobility
Ion mobility not required
Data compatibility
Inputs
Mass spectra
Outputs
Statistical results
Input formats
MGF, HDF5
Output formats
Python embedding matrices and task-dependent similarity or prediction arrays
Scope and considerations
Designed for
- Representation learning and similarity analysis of small-molecule tandem mass spectra.
Not designed for
- Raw chromatographic feature detection or definitive molecular identification from embedding similarity.
Important requirements
- The documented Python environment and pretrained model weights for local inference; MS/MS spectra with the required precursor metadata. Fine-tuning requires suitable labelled data and additional compute.
Considerations
- Learned embedding similarity is a computational relationship and does not by itself establish molecular identity.
Access and documentation
MIT-licensed code; pretrained weights and datasets are linked separately. The hosted demonstration is a separate execution route whose availability and submitted-data handling should be checked before use.
License: MIT
Evidence and review
Enveda intake rechecked against official README/documentation, repository access terms and primary publication metadata on 2026-10-27. Editorial source review only; no software execution or independent benchmarking. See docs/enveda-expansion-review-2026-10-27.md.
Core publications
Resource papers, direct updates and tool-specific methods. Selection policy.
- Self-supervised learning of molecular representations from millions of tandem mass spectra using DreaMS. Nature Biotechnology (2025). doi:10.1038/s41587-025-02663-3
Related tools
Entry history
Changes to this catalogue entry, not the tool's software release history. Acceptance and editorial review are separate.
- 2026-10-27 · First committed entry
Add 27 source-reviewed tools from Enveda intake
View change · Entry at this revision