pipette
ESEspañol

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

Jun A., Xiang Zhang, Xiaofan Zhang, Jiaqi Wei, Te Zhang, Yamin Deng, Pu Liu, Zongxiang Nie, Yiwen Yu, Yi Chen, Nanqing Dong, Zhiqiang Gao, Siqi Sun, Tiannan Guo

Peer-reviewed journalCode available

In the authors' words

This repository compiles the core resources used to construct the MassNet dataset, including: 1) FASTA sequence files for each species, used for database searching; 2) Standardized database search workflows based on FragPipe and Sage engines for unified processing of raw DDA-MS data and high-confidence peptide identification. Additionally, the repository provides the following data resources and supporting tools for downstream AI tasks: 1) Retention time (RT) prediction task: training and validation datasets constructed from FragPipe and Sage results, along with corresponding RT prediction model outputs; 2) Peptide-spectrum match (PSM) rescoring task: PSM datasets for training and evaluation results; Dataset construction tools: complete code and documentation for generating the above task-specific datasets. For detailed model training procedures and usage instructions, please refer to the following official repositories:DeepLC: https://github.com/CompOmics/DeepLCDDA-BERT: https://github.com/guomics-lab/DDA-BERT All resources provided in this repository enable full reproduction of the core experimental and analytical results reported in the manuscript "MassNet: billion-scale AI-ready mass spectrometry corpus enabling scalable deep Learning in proteomics".

Main resultThe abstract does not state a limitation.

Appeared: Friday, October 16. Nature Methods. Peer-reviewed journal.

DOI: 10.1038/s41592-026-03247-y