A Billion-Scale Dataset for Proteomics AI - Takeaways - MDSpire
Coming Soon: Introducing MDSpire News. Learn more
Conexiant’s news site is now MDSpire News. Learn more

A Billion-Scale Dataset for Proteomics AI

  • October 22, 2026

  • 3 min

Share

  • 1

    MassNet standardizes 1.54 billion tandem mass spectra from 648 projects, 27,643 runs, and 35 animal, plant, and microbial species.

  • 2

    Reanalysis with FragPipe and Sage yielded 548 million peptide-spectrum matches, 2.54 million unique peptides, and 133,422 proteins.

  • 3

    The MSDT format stores spectra with peptide identifications and experimental metadata, and its files were approximately 82% smaller than the source files.

  • 4

    Trained on 100 million peptide-spectrum matches, XuanjiNovo outperformed five other de novo sequencing models in a 15-species benchmark.

  • 5

    MassNet includes only spectra identified by FragPipe or Sage and does not systematically cover modifications such as phosphorylation, glycosylation, or ubiquitination.

Original Source(s)

Related Content