MultiMS2: A Multi-Modal, Multi-Energy MS2 Spectral Library

Overview

MultiMS2 is a curated mass spectrometry spectral library designed to address critical gaps in metabolomics research. This library provides high-quality MS/MS spectra across:

Multiple fragmentation methods: Collision-Induced Dissociation (CID) and Electron-Activated Dissociation (EAD)
Multiple collision energies: 20, 40, and 60 eV for CID; 12, 16, and 24 eV for EAD
Both ionization modes: Positive and negative

Why MultiMS2?

Confident metabolite identification relies on high-quality reference spectral libraries, yet most existing resources suffer from significant limitations:

Limited fragmentation diversity: Predominantly CID spectra, with EAD spectra remaining scarce despite their structural value
Restricted energy ranges: Single or limited collision energy coverage
Narrow acquisition conditions: Limited applicability across varied analytical workflows
Machine learning constraints: Insufficient diversity for training robust, generalizable models

MultiMS2 addresses these challenges by providing a curated resource that:

Enables reliable metabolite annotation across diverse experimental conditions
Supports development of generalizable machine learning models for spectrum prediction and structure elucidation
Facilitates comparative fragmentation studies between CID and EAD
Accelerates innovation in computational metabolomics

Data Availability

The complete dataset is publicly available through:

Zenodo: https://doi.org/10.5281/zenodo.17250693
MassIVE: https://doi.org/10.25345/C5GQ6RF85

Installation & Setup

Prerequisites

mzmine 3+
uv package manager
Docker (for initial conversion only)

Quick Start

Clone the repository

git clone https://github.com/yourusername/MultiMS2.git
cd MultiMS2

Download spectral data from Zenodo

uv run python notebooks/get_mzmls_from_zenodo.py

and then:

unzip "*.zip"

Configure mzmine batch files

Update the metadata file path in .mzmine/batch/*.mzbatch:

<parameter name="Database file">
    <current_file>/path/to/your/local/msmls_metadata_neg.tsv</current_file>
</parameter>

Usage

Spectral Extraction with mzmine¹²

The library generation uses mzmine batch processing for consistent, reproducible spectral extraction. Below are the commands for all library combinations:

NEXUS Library

Negative Mode:

# CID at different energies
mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_cid_20/*.mzML" \
  -o "scratch/nexus_neg_cid_20"

mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_cid_40/*.mzML" \
  -o "scratch/nexus_neg_cid_40"

mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_cid_60/*.mzML" \
  -o "scratch/nexus_neg_cid_60"

# EAD at different energies
mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_ead_12/*.mzML" \
  -o "scratch/nexus_neg_ead_12"

mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_ead_16/*.mzML" \
  -o "scratch/nexus_neg_ead_16"

mzmine -b ".mzmine/batch/nexus_library_generation_neg.mzbatch" \
  -i "scratch/nexus_mzml_centroided_neg_ead_24/*.mzML" \
  -o "scratch/nexus_neg_ead_24"

Positive Mode:

# CID at different energies
mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_cid_20/*.mzML" \
  -o "scratch/nexus_pos_cid_20"

mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_cid_40/*.mzML" \
  -o "scratch/nexus_pos_cid_40"

mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_cid_60/*.mzML" \
  -o "scratch/nexus_pos_cid_60"

# EAD at different energies
mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_ead_12/*.mzML" \
  -o "scratch/nexus_pos_ead_12"

mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_ead_16/*.mzML" \
  -o "scratch/nexus_pos_ead_16"

mzmine -b ".mzmine/batch/nexus_library_generation_pos.mzbatch" \
  -i "scratch/nexus_mzml_centroided_pos_ead_24/*.mzML" \
  -o "scratch/nexus_pos_ead_24"

Selleck Library

Negative Mode:

# CID energies
mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_cid_20/*.mzML" \
  -o "scratch/selleck_neg_cid_20"

mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_cid_40/*.mzML" \
  -o "scratch/selleck_neg_cid_40"

mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_cid_60/*.mzML" \
  -o "scratch/selleck_neg_cid_60"

# EAD energies
mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_ead_12/*.mzML" \
  -o "scratch/selleck_neg_ead_12"

mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_ead_16/*.mzML" \
  -o "scratch/selleck_neg_ead_16"

mzmine -b ".mzmine/batch/selleck_library_generation_neg.mzbatch" \
  -i "scratch/selleck_mzml_centroided_neg_ead_24/*.mzML" \
  -o "scratch/selleck_neg_ead_24"

Positive Mode:

# CID energies
mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_cid_20/*.mzML" \
  -o "scratch/selleck_pos_cid_20"

mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_cid_40/*.mzML" \
  -o "scratch/selleck_pos_cid_40"

mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_cid_60/*.mzML" \
  -o "scratch/selleck_pos_cid_60"

# EAD energies
mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_ead_12/*.mzML" \
  -o "scratch/selleck_pos_ead_12"

mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_ead_16/*.mzML" \
  -o "scratch/selleck_pos_ead_16"

mzmine -b ".mzmine/batch/selleck_library_generation_pos.mzbatch" \
  -i "scratch/selleck_mzml_centroided_pos_ead_24/*.mzML" \
  -o "scratch/selleck_pos_ead_24"

MSMLS Library

Negative Mode:

# CID energies
mzmine -b ".mzmine/batch/msmls_library_generation_neg.mzbatch" \
  -i "scratch/msmls_mzml_centroided_neg_cid_20/*.mzML" \
  -o "scratch/msmls_neg_cid_20"

mzmine -b ".mzmine/batch/msmls_library_generation_neg.mzbatch" \
  -i "scratch/msmls_mzml_centroided_neg_cid_40/*.mzML" \
  -o "scratch/msmls_neg_cid_40"

mzmine -b ".mzmine/batch/msmls_library_generation_neg.mzbatch" \
  -i "scratch/msmls_mzml_centroided_neg_cid_60/*.mzML" \
  -o "scratch/msmls_neg_cid_60"

# EAD energies (note: 12 eV not available)
mzmine -b ".mzmine/batch/msmls_library_generation_neg.mzbatch" \
  -i "scratch/msmls_mzml_centroided_neg_ead_16/*.mzML" \
  -o "scratch/msmls_neg_ead_16"

mzmine -b ".mzmine/batch/msmls_library_generation_neg.mzbatch" \
  -i "scratch/msmls_mzml_centroided_neg_ead_24/*.mzML" \
  -o "scratch/msmls_neg_ead_24"

Positive Mode:

# CID energies (note: 20 eV not available)
mzmine -b ".mzmine/batch/msmls_library_generation_pos.mzbatch" \
  -i "scratch/msmls_mzml_centroided_pos_cid_40/*.mzML" \
  -o "scratch/msmls_pos_cid_40"

mzmine -b ".mzmine/batch/msmls_library_generation_pos.mzbatch" \
  -i "scratch/msmls_mzml_centroided_pos_cid_60/*.mzML" \
  -o "scratch/msmls_pos_cid_60"

# EAD energies (note: 12 eV not available)
mzmine -b ".mzmine/batch/msmls_library_generation_pos.mzbatch" \
  -i "scratch/msmls_mzml_centroided_pos_ead_16/*.mzML" \
  -o "scratch/msmls_pos_ead_16"

mzmine -b ".mzmine/batch/msmls_library_generation_pos.mzbatch" \
  -i "scratch/msmls_mzml_centroided_pos_ead_24/*.mzML" \
  -o "scratch/msmls_pos_ead_24"

Performance Note: For optimal performance, consider copying files to a fast local disk before processing to avoid slow network I/O.

Quality Control and Validation

Metadata

Because of an issue during .mzML file conversion, COLLISION_ENERGY and FRAGMENTATION_METHOD are missing from negative CID files. To fix it, run:

uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/nexus_neg_cid_20_batch_library.mgf CID 20.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/nexus_neg_cid_40_batch_library.mgf CID 40.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/nexus_neg_cid_60_batch_library.mgf CID 60.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/selleck_neg_cid_20_batch_library.mgf CID 20.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/selleck_neg_cid_40_batch_library.mgf CID 40.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/selleck_neg_cid_60_batch_library.mgf CID 60.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/msmls_neg_cid_20_batch_library.mgf CID 20.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/msmls_neg_cid_40_batch_library.mgf CID 40.0
uv run python notebooks/edit_mgf_collision_fragmentation.py /Users/adrutz/Git/MultiMS2/scratch/msmls_neg_cid_60_batch_library.mgf CID 60.0

Concatenation

uv run python notebooks/concatenate_spectra.py

Modalities and Quality Filtering

After this, spectra from all sub-libraries and modalities are concatenated by running:

uv run python notebooks/filter_spectra_consistent.py

Only the spectra complying to the following rules are kept:

# Thresholds
min_precursor_height = 1000.0
min_precursor_purity = 0.9
min_signals = 3
min_explained_intensity = 0.4
min_explained_signals = 0.05
min_modalities = 2
min_intensity_ratio = 0.8
min_signals_ratio = 0.4

After charge consistency filter: 148888 spectra remain.
After min_precursor_height: 133624 spectra, 11998 unique (inchi_aux, adduct)
After min_precursor_purity: 128116 spectra, 11693 unique (inchi_aux, adduct)
After min_signals: 113652 spectra, 11346 unique (inchi_aux, adduct)
After min_explained_intensity: 67004 spectra, 9394 unique (inchi_aux, adduct)
After min_explained_signals: 66916 spectra, 9392 unique (inchi_aux, adduct)
After min_intensity_ratio: 61163 spectra, 9392 unique (inchi_aux, adduct)
After min_signals_ratio: 60359 spectra, 9392 unique (inchi_aux, adduct)
After min_explained_intensity: 60359 spectra, 9392 unique (inchi_aux, adduct)
After min_explained_signals: 60359 spectra, 9392 unique (inchi_aux, adduct)
After min_modalities per (inchi_aux, adduct): 3043 inchi_aux, 4563 unique (inchi_aux, adduct), 25126 unique (inchi_aux, adduct, modality)
Final spectra selected for output: 47630
Exported 47630 final spectra to scratch/filtered_spectra.mgf

(Both all and filtered MGF are exported)

Adduct Assignment Check

An additional check is performed to chemically validate numerically found adducts/losses. In other words, if a spectrum was recognized as [M-H2O+H]+, it checks if the compound contains hydroxyls, and so on. Checks are performed using RDKit³.

uv run python notebooks/validate_losses.py

From the 47,630 filtered spectra, 43,728 were validated and 3,902 discarded.

Metadata Consolidation

At this point, SELFIES⁴ can be added, metadata in the headers cleaned up, and unique feature IDs attributed using:

uv run python notebooks/consolidate_spectra.py --instrument_name ZENOTOF7600 --data_curator ARutz

MS-BUDDY⁵ Molecular Formula Annotation

MS-BUDDY provides molecular formula annotation to further structural validation:

uv run msbuddy \
-mgf "scratch/consolidated_spectra.mgf" \
-ms qtof \
-parallel \
-batch_size 100 \
-details \
-halogen \
-rel_int_denoise_cutoff 0 \
-top_n_per_50_da -1 \
-output "scratch/msbuddy"

Visualization

uv run python notebooks/viz_upset.py

In the end:

2,899 unique compounds were recorded in
- 4,210 unique compound-adduct modalities,
  - 17,170 unique compound-adduct-fragmentation modalities for a total of
    - 43,728 spectra

uv run python notebooks/viz_msbuddy.py

GNPS⁶ export

To export the TSV file required for GNPS libraries:

uv run python notebooks/convert_spectra_to_tsv.py --change_mzml_to_mzxml --split 10000

This should then pass the validation at https://gnps-quickstart.ucsd.edu/validatebatch

Data Processing Pipeline (here for reference)

Initial Data Conversion

The following steps document the complete data processing workflow from raw instrument files to curated spectral libraries.

Step 1: Convert `.wiff` to `.mzML`

Raw AB SCIEX .wiff files are converted to open-format .mzML using ProteoWizard⁷:

docker run -it --rm \
  -v .:/data \
  proteowizard/pwiz-skyline-i-agree-to-the-vendor-licenses \
  wine msconvert "*.wiff" \
  --ignoreUnknownInstrumentError

Step 2: Profile to Centroided Spectra

Profile mode spectra are converted to centroided format using CentroidR⁸:

files <- "/Volumes/T7/data/7600/ms2_libraries" |>
  list.files(pattern = ".mzML", recursive = TRUE, full.names = TRUE)

files |>
  purrr::walk(
    .f = CentroidR::centroid_one_file,
    pattern = "/profile/",
    replacement = "/centroided/"
  )

Note: These preprocessing steps have already been completed for the publicly available datasets on Zenodo and MassIVE.

Library Contents

Compound Collections

NEXUS: Diverse natural product and drug-like compounds
Selleck: Bioactive compound library focused on drug discovery
MSMLS: Metabolomics Standards Library compounds

Spectral Coverage

Collection	Ionization	Fragmentation	Energies Available
NEXUS	Positive	CID	20, 40, 60 eV
NEXUS	Positive	EAD	12, 16, 24 eV
NEXUS	Negative	CID	20, 40, 60 eV
NEXUS	Negative	EAD	12, 16, 24 eV
Selleck	Positive	CID	20, 40, 60 eV
Selleck	Positive	EAD	12, 16, 24 eV
Selleck	Negative	CID	20, 40, 60 eV
Selleck	Negative	EAD	12, 16, 24 eV
MSMLS	Positive	CID	40, 60 eV*
MSMLS	Positive	EAD	16, 24 eV*
MSMLS	Negative	CID	20, 40, 60 eV
MSMLS	Negative	EAD	16, 24 eV*

Some energy levels not available for certain MSMLS conditions

Applications

MultiMS2 enables:

Metabolite Annotation: High-confidence identification through multi-energy spectral matching
Machine Learning Development: Training data for spectrum prediction and structure elucidation models
Fragmentation Studies: Comparative analysis of CID vs. EAD fragmentation patterns
Method Development: Reference spectra for optimizing MS/MS acquisition parameters
Quality Assessment: Benchmarking datasets for evaluating annotation algorithms

Acknowledgments

This work was supported by a grant from the Swiss National Science Foundation (project MetabolinkAI, #10002786), and a grant from the Strategic Focal Area Personalized Health and Related Technologies (PHRT) of the ETH Domain (#603).

References

Schmid, R., Heuckeroth, S., Korf, A., Smirnov, A., Myers, O., Dyrlund, T. S., Bushuiev, R., Murray, K. J., Hoffmann, N., Lu, M., Sarvepalli, A., Zhang, Z., Fleischauer, M., Dührkop, K., Wesner, M., Hoogstra, S. J., Rudt, E., Mokshyna, O., Brungs, C., … Pluskal, T. (2023). Integrative analysis of multimodal mass spectrometry data in MZmine 3. Nature Biotechnology, 41(4), 447–449. https://doi.org/10.1038/s41587-023-01690-2 ↩
Brungs, C., Schmid, R., Heuckeroth, S., Mazumdar, A., Drexler, M., Šácha, P., Dorrestein, P. C., Petras, D., Nothias, L.-F., Veverka, V., Nencka, R., Kameník, Z., & Pluskal, T. (2025). MSnLib: efficient generation of open multi-stage fragmentation mass spectral libraries. Nature Methods, 22(10), 2028–2031. https://doi.org/10.1038/s41592-025-02813-0 ↩
Greg Landrum, Paolo Tosco, Brian Kelley, Ricardo Rodriguez, David Cosgrove, Riccardo Vianello, sriniker, Peter Gedeck, Gareth Jones, Eisuke Kawashima, NadineSchneider, Dan Nealschneider, Andrew Dalke, tadhurst-cdd, Matt Swain, Brian Cole, Samo Turk, Aleksandr Savelev, Alain Vaucher, … Juuso Lehtivarjo. (2025). rdkit/rdkit: 2025_09_1 (Q3 2025) Release (Version Release_2025_09_1). Zenodo. https://doi.org/10.5281/ZENODO.17232453 ↩
Krenn, M., Häse, F., Nigam, A., Friederich, P., & Aspuru-Guzik, A. (2020). Self-referencing embedded strings (SELFIES): A 100% robust molecular string representation. Machine Learning: Science and Technology, 1(4), 045024. https://doi.org/10.1088/2632-2153/aba947 ↩
Xing, S., Shen, S., Xu, B., Li, X., & Huan, T. (2023). BUDDY: molecular formula discovery via bottom-up MS/MS interrogation. Nature Methods, 20(6), 881–890. https://doi.org/10.1038/s41592-023-01850-x ↩
Wang, M., Carver, J. J., Phelan, V. V., Sanchez, L. M., Garg, N., Peng, Y., Nguyen, D. D., Watrous, J., Kapono, C. A., Luzzatto-Knaan, T., Porto, C., Bouslimani, A., Melnik, A. V., Meehan, M. J., Liu, W.-T., Crüsemann, M., Boudreau, P. D., Esquenazi, E., Sandoval-Calderón, M., … Bandeira, N. (2016). Sharing and community curation of mass spectrometry data with Global Natural Products Social Molecular Networking. Nature Biotechnology, 34(8), 828–837. https://doi.org/10.1038/nbt.3597 ↩
Chambers, M. C., Maclean, B., Burke, R., Amodei, D., Ruderman, D. L., Neumann, S., Gatto, L., Fischer, B., Pratt, B., Egertson, J., Hoff, K., Kessner, D., Tasman, N., Shulman, N., Frewen, B., Baker, T. A., Brusniak, M.-Y., Paulse, C., Creasy, D., … Mallick, P. (2012). A cross-platform toolkit for mass spectrometry and proteomics. Nature Biotechnology, 30(10), 918–920. https://doi.org/10.1038/nbt.2377 ↩
Rutz, A., & Rainer, J. (2025). CentroidR: Repository to centroid profile spectra. (Version 0.0.0.9001). Zenodo. https://doi.org/10.5281/ZENODO.17250308 ↩

Name		Name	Last commit message	Last commit date
Latest commit History 102 Commits
.github		.github
.mzmine/batch		.mzmine/batch
data		data
figures		figures
metadata		metadata
notebooks		notebooks
.gitignore		.gitignore
.zenodo.json		.zenodo.json
CITATION.cff		CITATION.cff
LICENSE		LICENSE
README.md		README.md
library_info.json		library_info.json
pyproject.toml		pyproject.toml
uv.lock		uv.lock

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

MultiMS2: A Multi-Modal, Multi-Energy MS2 Spectral Library

Overview

Why MultiMS2?

Data Availability

Installation & Setup

Prerequisites

Quick Start

Usage

Spectral Extraction with mzmine¹²

NEXUS Library

Selleck Library

MSMLS Library

Quality Control and Validation

Metadata

Concatenation

Modalities and Quality Filtering

Adduct Assignment Check

Metadata Consolidation

MS-BUDDY⁵ Molecular Formula Annotation

Visualization

GNPS⁶ export

Data Processing Pipeline (here for reference)

Initial Data Conversion

Step 1: Convert `.wiff` to `.mzML`

Step 2: Profile to Centroided Spectra

Library Contents

Compound Collections

Spectral Coverage

Applications

Acknowledgments

References

About

Uh oh!

Uh oh!

Languages

License

zamboni-lab/MultiMS2

Folders and files

Latest commit

History

Repository files navigation

MultiMS2: A Multi-Modal, Multi-Energy MS2 Spectral Library

Overview

Why MultiMS2?

Data Availability

Installation & Setup

Prerequisites

Quick Start

Usage

Spectral Extraction with mzmine12

NEXUS Library

Selleck Library

MSMLS Library

Quality Control and Validation

Metadata

Concatenation

Modalities and Quality Filtering

Adduct Assignment Check

Metadata Consolidation

MS-BUDDY5 Molecular Formula Annotation

Visualization

GNPS6 export

Data Processing Pipeline (here for reference)

Initial Data Conversion

Step 1: Convert .wiff to .mzML

Step 2: Profile to Centroided Spectra

Library Contents

Compound Collections

Spectral Coverage

Applications

Acknowledgments

References

Footnotes

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Uh oh!

Languages

Spectral Extraction with mzmine¹²

MS-BUDDY⁵ Molecular Formula Annotation

GNPS⁶ export

Step 1: Convert `.wiff` to `.mzML`