msmu: a computational framework for provenance-aware mass spectrometry-based proteomics analysis

Abstract

Heterogeneous data formats and analysis pipelines in computational proteomics hinder reproducibility, interoperability, and reuse of processed data. We present msmu, an open-source Python package that implements a modular, provenance-aware pipeline for post-search processing. Within a single MuData object, msmu links representations across peptide-spectrum match, peptide, and protein levels to machine-actionable provenance that records operations, parameter choices, software versions, and the execution environment. The object's provenance log directly enables workflow replay with content-hash verification, without separately maintained scripts or provenance records. We demonstrate cross-platform reproducibility between macOS and Debian; replay reproduces the original protein inference map exactly (Jaccard indices = 1), with protein q-values and intensities agreeing to machine precision (root-mean-square deviation [≤] 2.0X10-17; maximum absolute difference [≤] 1.8X10-15). Furthermore, the same data structure supports seamless integration of proteomics with other MuData-compatible modalities. By coupling proteomics data with verifiable provenance in a single object, msmu advances FAIR-aligned reuse and multimodal interoperability.