Register for research access
Please use your institutional email address where possible.
Already registered? Log in.
Please use your institutional email address where possible.
Already registered? Log in.
MAGNET (Mass-spectrum Annotation via Gated-intensity Networks) is an encoder–decoder transformer that reads an electron-ionisation GC-MS spectrum and proposes the molecular structure as SMILES. Peaks enter the encoder as intensity-gated (m/z, intensity) tokens; the decoder is sampled many times and the structures are ranked by consensus frequency — how often each unique molecule is drawn. An optional molecular formula floats formula-matching candidates to the top.
The model is co-trained on simulated and real-world GC-MS data.
If MAGNET helps your research, please cite:
Please choose the contact that best matches your request.
For access problems, account administration, service availability, or requests to access or delete your personal data.
info@nxt-spectra.deFor questions about MAGNET's methodology, molecular predictions, research use, the manuscript, or scientific collaboration.
ajnabiul.hoque@uni-muenster.deGlorius Group · Organic Chemistry Institute · University of Münster · Corrensstraße 36, 48149 Münster, Germany