Skip to content
Open access

MassNet: billion-scale AI-ready mass spectral corpus empowering scalable deep learning in proteomics

Oct 2026 · Nature Methods · Vol 23, pp. 2206 - 2218 · 0 citations · 53 references
Medicine

TL;DR

Building on the basis of MassNet, XuanjiNovo is a nonautoregressive transformer for de novo peptide sequencing that substantially outperforms previous methods, marking a critical milestone for AI proteomics research.

Abstract

Here we present MassNet, inspired by large-scale data corpora of text and images; MassNet is an artificial intelligence (AI)-ready data corpus for mass spectrometry (MS)-based proteomics. The corpus contains 1.5 billion tandem MS spectra from high-quality data-dependent acquisition (DDA) MS data across 35 species. Approximately 1.7 million peptide precursors are from human samples. We further present an AI-ready DDA data format called Mass Spectrometry DDA Tensor (MSDT) for compact and standardized storage, curation and reuse of data from multiple sources, which can be optionally converted to a Lightning Memory-Mapped Database for accelerated input/output in deep learning. On the basis of MassNet, we develop XuanjiNovo, a nonautoregressive transformer model that leverages a curriculum learning strategy for de novo peptide sequencing. XuanjiNovo achieves 38.35–143.57% improvement over state-of-the-art methods on human Astral data. MassNet also supports additional downstream tasks, including retention time prediction and PSM rescoring. MassNet represents a foundational MS-based proteomics data resource, marking a critical milestone for AI proteomics research. This Article presents MassNet, a large-scale AI-ready corpus of 1.5 billion MS/MS spectra spanning 35 species, paired with a standardized storage format called Mass Spectrometry DDA Tensor (MSDT). Built on this resource, XuanjiNovo is a nonautoregressive transformer for de novo peptide sequencing that substantially outperforms previous methods.

Read PDF

Similar papers

Open access Sep 2026

Learning from tandem mass spectra at scale with a self-supervised foundation model for proteomics

InstaNovo-FM is presented, a self-supervised foundation model for bottom-up proteomics trained to reconstruct masked regions of tandem mass spectra and establishes a unified representation space for peptide fragmentation spectra, enabling robust transferability across the proteomics ecosystem.

M. Nieuwoudt, Marco Reverenna, Divanisha Patel et al. · 1 citation · ⚡1
Open access Oct 2026

Pretrained spectrum representations for diverse tandem mass spectrometry proteomics tasks.

This work demonstrates that a pretrained spectrum encoder for tandem mass spectrometry proteomics trained on de novo sequencing learns generalizable representations of spectra, improves performance on downstream tasks where training data is limited, and can ultimately enhance data acquisition and analysis in proteomics...

Justin A. Sanders, Melih Yilmaz, Jacob H. Russell et al. · 0 citations
Open access Sep 2026

Prioritizing Peptides for Targeted Mass Spectrometry Experiments Using Deep Learning.

Bromo is introduced, a transformer-based deep learning model that ranks peptide precursors from a given protein by their relative response, taking charge state into account, and consistently outperforms existing sequence-based methods across diverse, independent data sets.

Shreyash Sonthalia, Bo Wen, Priank Dasgupta et al. · 0 citations
2027

ProteoRift Protocols: A Deep Learning Filtering Workflow for Peptide Identification in Mass Spectrometry Data.

This chapter presents a machine-learning model, called ProteoRift, capable of predicting multiple properties of the peptide directly from the spectra, which helps to apply more targeted filters in the database search and provides up to 12× speedup as compared to mass-only filtering in peptide identification.

Bilal Shabbir, U. Tariq, Fahad Saeed · 0 citations
2027

Annotating Tandem Mass Spectra in Metabolomics with DreaMS and MassSpecGym.

Overall, this chapter aims to empower biologists and chemists to harness machine learning for metabolomics, transforming unannotated spectra into interpretable chemical knowledge.

Filip Jozefov, Roman Bushuiev, Anton Bushuiev et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.