Editorial additions pending. This submission has been accepted into the catalogue. Detailed metadata and editorial review may follow in later updates. Missing information does not mean a capability is unsupported.
Overview
ICEBERG predicts tandem mass spectra from candidate molecular structures using a learned fragment generator and an intensity predictor. Predicted spectra can be compared with an experimental spectrum to rank supplied candidates. The ms-pred repository also contains separate models; this entry covers ICEBERG, not every method in that repository.
What can this tool do?
Common uses
- Predict MS/MS spectra for candidate structures.
- Rank a defined candidate set by comparison with an experimental spectrum.
Scientific fit
Primary function
Metabolite annotation and identification
Other functions
Machine learning and prediction
Data compatibility
Inputs
Molecular structures, Mass spectra
Outputs
Mass spectra, Compound list or annotations
Input formats
Other, SMILES candidate structures; experimental spectra for candidate ranking.
Output formats
Other, Predicted spectra and candidate rankings; model-specific outputs.
Scope and considerations
Designed for
- Structure-to-spectrum prediction and candidate ranking for small molecules.
Not designed for
- Raw chromatographic feature extraction or guaranteed de novo identification from a spectrum.
Important requirements
- Use matching model weights, molecular-structure inputs and model-specific acquisition metadata. Follow the prediction example for the chosen ICEBERG version.
Considerations
- Candidate ranking is limited by the supplied candidate set and model applicability.
- The repository distinguishes ICEBERG 1.0, 2.0 and 2.1. Record the model and weights used.
- Training hardware requirements must not be presented as the minimum requirement for pretrained inference.
- Software execution and prediction performance were not tested during this editorial preparation.
Education
Learn this tool
External tutorials and workshops collected by Navigator. Entries marked Needs review have not been checked end to end.
Access and documentation
Python source, model instructions and links to pretrained weights are provided in the repository. Dataset access and pretrained-model use are distinct from training a new model; consult their respective terms and requirements.
Evidence and review
Draft editorial additions checked against the official ms-pred README and direct ICEBERG publications. Original developer-submission attribution is retained; no additional developer verification is claimed.
Core publications
Resource papers, direct updates and tool-specific methods. Selection policy.
- Goldman S, Li J, Coley CW. Generating Molecular Fragmentation Graphs with Autoregressive Neural Networks. Analytical Chemistry 96, 3419-3428 (2024). doi:10.1021/acs.analchem.3c04654
Original ICEBERG 1.0 method.
- Preprint · not peer reviewed Wang R, et al. Neural Spectral Prediction for Structure Elucidation with Tandem Mass Spectrometry. bioRxiv (2025). Preprint. doi:10.1101/2025.05.28.656653
ICEBERG 2.0 citation listed by the official repository; later repository updates should be identified by model version.
Related tools
Entry history
Changes to this catalogue entry, not the tool's software release history. Acceptance and editorial review are separate.
Original submission and discussion
- 2026-09-15 · Entry update
Keep ICEBERG output-format label concise
View change · Entry at this revision - 2026-09-15 · Entry update
Draft ICEBERG model scope and version-specific citations
View change · Entry at this revision - 2026-09-14 · First committed entry
Add draft tool entry for ICEBERG
View change · Entry at this revision