# MSnLib: Multi-Stage (MSn) Tandem Mass Spectral Libraries of 4.5 Million Orbitrap Spectra (IOCB Prague, Parquet)

## Facts

- Provider: Corinna Brungs, Robin Schmid, Tomas Pluskal et al. (IOCB Prague)
- License: CC-BY-4.0 (https://zenodo.org/records/21105617)
- Size: 125 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/structure-epflai/msnlib
- Access: public, readable without an account
- DOI: https://doi.org/10.5281/zenodo.21105617
- Category: Chemistry & Materials
- Source: https://doi.org/10.1038/s41592-025-02813-0
- Page: https://scigantic.com/archive/ba03403f-84b8-5a16-877a-aa8dbddd8e9f
- Keywords: Chemistry & Materials, spectroscopy, benchmark, Parquet, Hugging Face

## About

MSnLib is an open library of multi-stage fragmentation (MSn) tandem mass spectra for small molecules, built by the Pluskal lab at IOCB Prague by flow-injection analysis of commercial compound libraries on an Orbitrap ID-X and automatic library generation in mzmine. Each compound ion carries MS2 spectra, pseudo-MS2 spectra (a full MSn tree merged into one spectrum) and the individual MS3+ fragmentation stages, in positive and negative ionization modes. This Hugging Face copy holds the 4,532,018 MGF records of Zenodo record 21105617 as Parquet with the original peak order and intensities, precursor m/z, adduct, collision energy and source metadata kept as text, plus a compounds table of the nine detected libraries; MS2 and MSn exports overlap by design, so merged spectra are not independent measurements. Use it to train and benchmark spectrum-to-structure models, spectral similarity search and fragmentation-tree prediction for metabolomics and natural products; load with pandas, pyarrow or datasets. License CC BY 4.0 (Zenodo, DOI 10.5281/zenodo.21105617). Cite Brungs, Schmid, Heuckeroth et al., MSnLib: efficient generation of open multi-stage fragmentation mass spectral libraries, Nature Methods 22, 2028 (2025).

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("structure-epflai/msnlib")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `data/compounds/20250828_9libraries_only_detected_cleaned.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/ba03403f-84b8-5a16-877a-aa8dbddd8e9f
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/ba03403f-84b8-5a16-877a-aa8dbddd8e9f
- Over MCP: `get_schema_card` with id `ba03403f-84b8-5a16-877a-aa8dbddd8e9f` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Chemistry & Materials datasets](https://scigantic.com/topics/chemistry-and-materials)
- [Spectroscopy datasets](https://scigantic.com/topics/spectroscopy)
