Metabolomics Navigator Tools, strategies & workflow guidance for metabolomics

Spectral processing and comparison

MassFormer

A research implementation of a graph-transformer model for predicting small-molecule tandem mass spectra from molecular structures.

Platform-independentUntargeted metabolomicsPython packageCommand lineOpen source
Access toolSuggest an update

Overview

MassFormer predicts small-molecule tandem mass spectra from molecular structures with a graph-transformer model. The research implementation includes inference scripts, pretrained checkpoints and training/evaluation configurations; prediction takes candidate structures rather than an unknown experimental spectrum alone.

What can this tool do?

In-silico spectrum prediction

Common uses

  • Predict MS/MS spectra for small molecules from structure.
  • Evaluate or benchmark learned spectrum-prediction models.

Scientific fit

Primary function

Spectral processing and comparison

Other functions

Machine learning and prediction

Analysis types

Untargeted metabolomics

MS level

MS/MS

Acquisition

Targeted MS/MS, General or unspecified acquisition

Ion mobility

Not applicable

Data compatibility

Inputs

Molecular structures

Outputs

Mass spectra

Input formats

CSV containing molecular identifiers and SMILES

Output formats

CSV of predicted spectra and associated metadata

Scope and considerations

Designed for

  • Research and benchmarking of learned small-molecule MS/MS spectrum prediction.

Not designed for

  • De novo recovery of a molecular structure from an experimental spectrum or automatic identification without candidate comparison.

Important requirements

  • A Unix-like Python/PyTorch environment, model checkpoint and molecular SMILES. GPU acceleration is recommended for larger jobs; documented CPU inference is also available.

Considerations

  • Predictions depend on molecular and collision-condition coverage; unsupported or unparseable SMILES can be omitted by the inference script.

Access and documentation

Primary accessDocumentation

BSD-2-Clause code. Public demo data and checkpoints are provided separately; reproducing NIST-based experiments additionally requires separately licensed NIST data.

License: BSD-2-Clause

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.

  1. Tandem mass spectrum prediction for small molecules using graph transformers. Nature Machine Intelligence (2024). doi:10.1038/s42256-024-00816-8

Related tools

BLINKSimilar function · A Python package for efficient cosine-based similarity scoring and matching of large collections of fragmentation mass spectra, including analog-search workflows.DreaMSSimilar function · A transformer-based framework for learning molecular representations from tandem mass spectra, supporting spectral similarity, clustering, molecular networking, and downstream prediction tasks.FastEISimilar function · Search electron-ionization spectra against a large predicted reference library.FragHubSimilar function · Standardize and combine reference mass-spectral libraries.

Entry history

Changes to this catalogue entry, not the tool's software release history. Acceptance and editorial review are separate.

  1. 2026-10-27 · First committed entry

    Add 27 source-reviewed tools from Enveda intake

    View change · Entry at this revision

Full GitHub history