ICEBERG Spectrum Retrieval (PubChem Atlas)

Under development. URL might change in the future. Please report any issues to runzhong@mit.edu

Query spectra

Paste one or more MS/MS spectra below. For multiple collision energies, click “Add another spectrum” and select the corresponding NCE (5–100%). The NCE will be converted to absolute eV using the precursor m/z computed from the formula and adduct.

If provided, spectra in the .ms file (one block per collision energy) will be parsed and used as query spectra. You can still edit the collision energy and peaks below.

Use spectra of a phospholipid standard (C27H56NO7P).

Spectrum 1
Helps remove electronic noise before computing entropy similarity.

Download predicted spectra (.mgf / .msp)

Download ICEBERG-predicted spectra for a specific PubChem formula from the atlas. You can either download the raw MGF only, or download MGF and a converted MSP file.
All NIST structures are excluded to avoid license conflict. You can import the predicted library into your NIST MS Search software.

The current ICEBERG PubChem atlas contains 91,424,208 unique structures from PubChem.

Please cite:
Neural Spectral Prediction for Structure Elucidation with Tandem Mass Spectrometry.
Runzhong Wang, Mrunali Manjrekar, Babak Mahjour, Julian Avila-Pacheco, Joules Provenzano, Erin Reynolds, Magdalena Lederbauer, Eivgeni Mashin, Samuel Goldman, Mingxun Wang, Jing-Ke Weng, Desirée L. Plata, Clary B. Clish, Connor W. Coley. bioRxiv 2025.
View the bioRxiv preprint

ICEBERG code: https://github.com/coleygroup/ms-pred

About ICEBERG

ICEBERG is a neural network model that predicts mass spectra from molecular structures. It can be used to elucidate unknown structures given mass spec.

ICEBERG model illustration
Coley Lab logo MIT Coley Group: https://coley.mit.edu/
Copied!
Running retrieval
Starting…