Expanding the mzPeak tool suite for the v1.0 specification

[tivdnbos]

Title

Expanding the mzPeak tool suite around the v1.0 spec format

Abstract

mzPeak is an open format for mass spectrometry data, developed in collaboration with HUPO-PSI as a successor to mzML for high-throughput and multidimensional workflows. It separates large numerical signal data from structured, human-readable metadata, stores signal data in Apache Parquet, and remains compatible with PSI controlled vocabularies. In benchmarks, converted files range from 0.18x to 0.57x the size of the equivalent mzML file, depending on instrument type, and a single spectrum can be read from a multi-gigabyte remote file in under two seconds.

mzPeak v1.0 is planned to be frozen and submitted to the HUPO-PSI Document Process in October 2026, ahead of EuBIC 2027. The specification will therefore remain fixed during the hackathon, while several implementations and supporting tools still have gaps that can be addressed during the event.

The hackathon will focus on extending this tool suite. Participants can add mzPeak support to existing proteomics software, including search engines and viewers, or work on the standalone implementations, converters, and validator. All contributions can be tested against the existing example files and the frozen v1.0 specification.

Project Plan

High-level plan:

• Identify gaps: inventory which language implementations, readers, writers, and converters are missing or incomplete against the frozen v1.0 specification, and identify participants who maintain existing proteomics software that could add mzPeak support.
• Build: teams implement or extend a standalone reader, writer, converter, or language binding, or add mzPeak support to a proteomics tool they already maintain.
• Evaluate: test each contribution against the mzPeak example files and mzPeakValidator, and test read/write or conversion consistency where relevant.
• Disseminate: merge completed and tested contributions into the relevant repositories and add new implementations and integrations to mzpeak.org.

During the hackathon:

1. Walk through the frozen v1.0 specification and the current status of each implementation: Rust (read/write reference), Python and R (read-only), C# (read/write), JavaScript/TypeScript (read-only), C++ (read-only, early stage), and Java (read/write demonstrator).
2. Participants select a task: add write support to the Python or R implementations, extend or harden the C++ or Java implementations, improve converter coverage, add validator checks, or add mzPeak read/write support directly to proteomics software they maintain.
3. Build and test against the shared example files (small.mzpeak, small.chunked.mzpeak, small.numpress.mzpeak, Example_Processed.img.mzpeak, has_uv.mzpeak) aand, where possible, against a public ProteomeXchange run converted during the event.
4. Cross-check contributions for read/write consistency and validator compliance against the frozen specification.
5. Hold a closing session in which each team presents what it built, what is ready to merge, and which existing tools now read or write mzPeak.

After the hackathon:

• Merge viable contributions into the relevant HUPO-PSI/mzPeak repositories and list new implementations and integrations on mzpeak.org.
• Track existing proteomics software that gained mzPeak support during the event and document these implementations for the PSI review.
• Assign remaining implementation gaps to the relevant maintainers for follow-up during the PSI review process.
• Use the expanded tool suite and new software integrations as implementation evidence during the PSI Document Process and in the ASMS 2027 presentation.

Technical Details

Programming language: open by design and dependent on the task selected by each participant. Existing mzPeak implementations span Rust, Python, R, C#, JavaScript/TypeScript, C++, and Java, with different levels of read/write support and maturity.

Will build on existing software: the mzPeak reference implementation and specification, existing language-specific implementations, mzML2mzPeak, mzPeakValidator, Apache Arrow/Parquet tooling, and proteomics software maintained by participating developers.

Datasets: mzPeak's published example files, plus a small set of public ProteomeXchange/PRIDE runs converted during the event. Specific PXD accessions will be selected before the hackathon.

Contact information

Tim Van Den Bossche, CompOmics, VIB - Ghent University, tim.vandenbossche@ugent.be (confirmed)
Sam Wein, University of Tübingen / OpenMS (to be confirmed)
Oliver Kohlbacher, University of Tübingen (to be confirmed)
Joshua Klein (remote afternoon sessions, to be confirmed)