Pretrained spectrum representations for diverse tandem mass spectrometry proteomics tasks
5 affiliations
- 1Paul G. Allen School of Computer Science and Engineering, University of Washington, Seattle, WA, 98195, USA
- 2Dept. of Chemistry, University of Washington, Seattle, WA, 98195, USA
- 3Department of Computer Science, University of Antwerp, Antwerp, 2020, Belgium
- 4Talus Bioscience, Seattle, WA, USA, 91895
- 5Department of Genome Sciences, University of Washington, Seattle, WA, USA, 98195; Paul G. Allen School of Computer Science and Engineering, University of Washington, Seattle, WA, 98195, USA. Electronic address: [email protected]
Abstract
Mass spectrometry is the dominant technology in the field of proteomics, enabling high-throughput analysis of the protein content of complex biological samples. Due to the complexity of the instrumentation and resulting data, sophisticated computational methods are required for the processing and interpretation of acquired mass spectra. Machine learning has shown great promise to improve the analysis of mass spectrometry data, with numerous purpose-built methods for improving specific steps in the data acquisition and analysis pipeline reaching widespread adoption. Each of these tasks rely on a shared fundamental understanding of the information present in a mass spectrum. Here, we propose a transfer learning framework, CasanovoTL, which utilizes pretrained spectrum representations, derived from a de novo sequencing model trained on a massive proteomics dataset, rather than training a spectrum encoder from scratch for each task. We show that using these pre-trained spectrum representations significantly improves our performance on the four downstream tasks of spectrum quality prediction, chimericity prediction, phosphorylation prediction, and glycosylation status prediction. Finally, we perform end-to-end finetuning, as well as multi-task training, and find that finetuned spectrum representations further improve performance on each individual task. Overall, our 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 experiments.
Related
- Tesorai Search: Cloud-based Database Search Engine Boosts Identifications for Mass Spectrometry Proteomics With a Pretrained Peptide-spectrum Deep-learning Model
- Mass Spectrometry-Based Structural Proteomics and Analytical Ultracentrifugation Characterize the Structures of Oligomeric Apolipoprotein E4 and its Variants
- Cross-Platform Evaluation Reveals Sample Preparation-Driven Biological Biases despite Advances in Mass Spectrometry Performance in Serum Proteomics.
- Data processing in proteomics and metabolomics: liquid chromatography-mass spectrometry-based omics with divergent computational paradigms.
- Foundation model enables interpretable open and error-tolerant searching for mass spectrometry-based proteomics
BibTeX
@article{sanders2026pretrained,
title = {{Pretrained spectrum representations for diverse tandem mass spectrometry proteomics tasks}},
author = {Justin Sanders and Melih Yilmaz and Jacob H. Russell and Wout Bittremieux and William E. Fondrie and Nicholas McIlvain Riley and Sewoong Oh and William Stafford Noble},
journal = {Molecular \& Cellular Proteomics},
year = {2026},
month = oct,
doi = {10.1016/j.mcpro.2026.101673},
url = {https://linkinghub.elsevier.com/retrieve/pii/S1535947626001696},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Pretrained spectrum representations for diverse tandem mass spectrometry proteomics tasks
AU - Justin Sanders
AU - Melih Yilmaz
AU - Jacob H. Russell
AU - Wout Bittremieux
AU - William E. Fondrie
AU - Nicholas McIlvain Riley
AU - Sewoong Oh
AU - William Stafford Noble
PY - 2026
DA - 2026/10/01
JO - Molecular & Cellular Proteomics
DO - 10.1016/j.mcpro.2026.101673
UR - https://linkinghub.elsevier.com/retrieve/pii/S1535947626001696
AB - Mass spectrometry is the dominant technology in the field of proteomics, enabling high-throughput analysis of the protein content of complex biological samples. Due to the complexity of the instrumentation and resulting data, sophisticated computational methods are required for the processing and interpretation of acquired mass spectra. Machine learning has shown great promise to improve the analysis of mass spectrometry data, with numerous purpose-built methods for improving specific steps in the data acquisition and analysis pipeline reaching widespread adoption. Each of these tasks rely on a shared fundamental understanding of the information present in a mass spectrum. Here, we propose a transfer learning framework, CasanovoTL, which utilizes pretrained spectrum representations, derived from a de novo sequencing model trained on a massive proteomics dataset, rather than training a spectrum encoder from scratch for each task. We show that using these pre-trained spectrum representations significantly improves our performance on the four downstream tasks of spectrum quality prediction, chimericity prediction, phosphorylation prediction, and glycosylation status prediction. Finally, we perform end-to-end finetuning, as well as multi-task training, and find that finetuned spectrum representations further improve performance on each individual task. Overall, our 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 experiments.
SN - 1535-9476
SN - 1535-9484
ER -