Why mzPeak

mzPeak stores mass spectra as Apache Parquet tables inside a ZIP archive, alongside a small JSON index. It is designed as a modern successor to mzML: a columnar, analytics-native layout that is dramatically smaller on disk, supports partial reads directly over HTTP, and maps losslessly onto the HUPO-PSI mzML / PSI-MS controlled vocabulary. The format is governed by HUPO-PSI and is currently at v0.9, Draft 5 — a working draft, not yet ratified through the HUPO-PSI Document Process; standardization is proceeding via HUPO-PSI from 2026. This page is meant to help you evaluate mzPeak honestly, get hands-on quickly, and cite the work.

mzPeak vs mzML

Dimension	mzML	mzPeak
Typical file size	XML text; frequently inflates past the vendor RAW size	0.18×–0.57× of the equivalent mzML (average ~0.37×), losslessly — and smaller than the vendor RAW file too
Random access / partial reads over HTTP	Indexed, but generally read as a whole file	Read a single spectrum from a multi-GB file over HTTP range requests, without downloading the whole file
Columnar analytics (Arrow / Pandas / Polars)	Not columnar; requires parsing into other structures	Native — Parquet reads directly into Apache Arrow, Pandas, or Polars
Metadata model (PSI-MS CV)	HUPO-PSI mzML / PSI-MS controlled vocabulary	Aligns with the same PSI-MS controlled vocabulary, so mzML↔mzPeak conversion is lossless
Imaging (MSI) support	Handled separately (imzML)	Spectra, chromatograms, and ion images for MS-imaging in one format
Streaming write maturity	Mature, universal	Rust and C# read+write; Java read+write (proof-of-concept); other languages read-only so far
Standardization status	Ratified, universal HUPO-PSI standard	v0.9, Draft 5 working draft; HUPO-PSI standardization from 2026
Tool ecosystem breadth	Decades of tools across the field	Growing — libraries in seven languages, in-browser viewer, validator, and an mzML converter

Implementation maturity

All implementations are built on Apache Arrow / Parquet.

Language	Capability
Rust	Reference implementation, read + write
C#	Read + write
Java	Read + write (proof-of-concept)
Python	Read
R	Read
JavaScript / TypeScript	Read
C++	Read

What mzML still does better today

mzML is a ratified, universal HUPO-PSI standard with decades of tool support across essentially the whole field — it will read anywhere. mzPeak is a v0.9 working draft: most of its libraries are read-only so far, and there is not yet a direct vendor-RAW converter (see the roadmap note below). If you need maximum compatibility with existing pipelines right now, mzML remains the safe default; mzPeak is the choice when file size, partial HTTP access, or columnar analytics matter.

Quickstart

1. Grab a .mzpeak file. The public corpus has ~79 real datasets, each mzPeak alongside its original, at data.mzpeak.org/v09.

2. Open it in the browser viewer. Go to the mzPeak Viewer and drag-and-drop your file, or use "Open demo" — no install, no upload. It streams over HTTP range requests and shows spectra and chromatograms for LC-/GC-MS, and ion images for MS-imaging.

3. Validate it with the mzPeak Validator — either in the browser (upload up to 1 GB, or paste an HTTPS/s3:// URL) or with the CLI:

   mzpeak-validate file.mzpeak

4. Make your own by converting an existing mzML with mzML2mzPeak. Because both formats share the PSI-MS controlled vocabulary, this conversion is lossless. See the repository for build and usage details.

A direct vendor RAW → mzPeak path (via ProteoWizard msconvert) is on the roadmap, not yet available. For now, convert vendor RAW to mzML first, then use mzML2mzPeak.

How to cite

If you use mzPeak in your work, please cite:

Tim Van Den Bossche, Theodore Alexandrov, Aivett Bilbao, Wout Bittremieux, Federico Ivan Brigante, Matthew Chase Chambers, Joshua Charkow, Eric Deutsch, Andrew W. Dowsey, Yasin El Abiead, Ralf Gabriels, Helge Hecht, Steffen Heuckeroth, Joshua A. Klein, Michael Knierman, Lennart Martens, Robert L. Moritz, Laura-Isobel McCall, Steffen Neumann, Yasset Perez-Riverol, Hannes L. Röst, Elliott J. Price, Jim Shofstahl, David L. Tabb, Julian Uszkoreit, Juan Antonio Vizcaíno, Mingxun Wang, Sander Willems, Dirk Winkelhardt, Oliver Kohlbacher, Samuel P. Wein. "mzPeak: Designing a Scalable, Interoperable, and Future-Ready Mass Spectrometry Data Format." Journal of Proteome Research, 2025, 24(11), 5329–5335. DOI: 10.1021/acs.jproteome.5c00435.

@article{vandenbossche2025mzpeak,
  author  = {Tim Van Den Bossche and Theodore Alexandrov and Aivett Bilbao and Wout Bittremieux and Federico Ivan Brigante and Matthew Chase Chambers and Joshua Charkow and Eric Deutsch and Andrew W. Dowsey and Yasin El Abiead and Ralf Gabriels and Helge Hecht and Steffen Heuckeroth and Joshua A. Klein and Michael Knierman and Lennart Martens and Robert L. Moritz and Laura-Isobel McCall and Steffen Neumann and Yasset Perez-Riverol and Hannes L. R{\"o}st and Elliott J. Price and Jim Shofstahl and David L. Tabb and Julian Uszkoreit and Juan Antonio Vizca{\'i}no and Mingxun Wang and Sander Willems and Dirk Winkelhardt and Oliver Kohlbacher and Samuel P. Wein},
  title   = {mzPeak: Designing a Scalable, Interoperable, and Future-Ready Mass Spectrometry Data Format},
  journal = {Journal of Proteome Research},
  year    = {2025},
  volume  = {24},
  number  = {11},
  pages   = {5329--5335},
  doi     = {10.1021/acs.jproteome.5c00435},
  url     = {https://pubs.acs.org/doi/full/10.1021/acs.jproteome.5c00435}
}

The mzPeak specification is currently licensed under CC-BY-ND 4.0.