Self-supervised learning of molecular representations from millions of tandem mass spectra using DreaMS

2025/05/23 by Roman Bushuiev, Anton Bushuiev, Raman Samusevich +3 · 3 voices · 59 citations
Biochemistry, Genetics and Molecular Biology · Chemistry · Computer Science · Engineering · #Advanced Chemical Sensor Technologies #Annotation #Artificial intelligence #Artificial neural network #Biological system #Biology #Chemistry #Chromatography #Computational Drug Discovery Methods #Computer science #Machine learning #Mass spectrometry #Mass spectrum #Materials science #Metabolomics and Mass Spectrometry Studies #Pattern recognition (psychology) #Physics #Spectral line #Tandem #Tandem mass spectrometry

paper · pdf · doi:10.1038/s41587-025-02663-3

published in Nature Biotechnology 44(4), 630-640 (Nature Portfolio)

openalex publication_date 2025/05/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06

Abstract

Characterizing biological and environmental samples at a molecular level primarily uses tandem mass spectroscopy (MS/MS), yet the interpretation of tandem mass spectra from untargeted metabolomics experiments remains a challenge. Existing computational methods for predictions from mass spectra rely on limited spectral libraries and on hard-coded human expertise. Here we introduce a transformer-based neural network pre-trained in a self-supervised way on millions of unannotated tandem mass spectra from our GNPS Experimental Mass Spectra (GeMS) dataset mined from the MassIVE GNPS repository. We show that pre-training our model to predict masked spectral peaks and chromatographic retention orders leads to the emergence of rich representations of molecular structures, which we named Deep Representations Empowering the Annotation of Mass Spectra (DreaMS). Further fine-tuning the neural network yields state-of-the-art performance across a variety of tasks. We make our new dataset and model available to the community and release the DreaMS Atlas-a molecular network of 201 million MS/MS spectra constructed using DreaMS annotations.

Citations

• Textbooks Are All You Need
• On the Opportunities and Risks of Foundation Models
• Language Models are Few-Shot Learners
• PyTorch: An Imperative Style, High-Performance Deep Learning Library
• Attention Is All You Need
• Colorful Image Colorization

Cited by

• Agent-Guided Relational Concept Discovery: Toward Interpretable Surgical Margin Assessment
• Conformal Graph Prediction with Z-Gromov-Wasserstein Distances
• MS-GPT: Rethinking MS/MS De Novo Structure Elucidation as Spectrum-Induced Posterior Querying of a Molecule-Language Model
• LSM-MS2: A Foundation Model Bridging Spectral Identification and Biological Interpretation
• End-to-end deep attention-based multitask pipeline for predicting uncertainty-quantified peptide properties from mass spectrometry data
• New technologies for improving the accuracy of neuroimaging: Stochastic Gaussian-masked denoiser enhances subcellular structures and neural dynamics
• MARLIN: De Novo Molecular Structure Elucidation from Tandem Mass Spectra without a Ground-Truth Formula
• MSAlign: Aligning Molecule and Mass Spectra Foundation Models for Metabolite Identification
• Physics-informed deep learning for plasmonic sensing of nanoscale protein dynamics in solution. [europepmc]
• On the use of TabPFN on mass spectrometry analysis of volatile organic compounds. [europepmc]
• Perspectives in computational mass spectrometry: recent developments and key challenges. [europepmc]
• A Perspective on Unintentional Fragments and Their Impact on the Dark Metabolome, Untargeted Profiling, Molecular Networking, Public Data, and Repository Scale Analysis. [europepmc]
• Cross ionization mode chemical similarity prediction between tandem mass spectra in metabolomics. [europepmc]
• Learning from All Views: A Multiview Contrastive Framework for Metabolite Annotation. [europepmc]
• Artificial intelligence to investigate metabolomics data for precision medicine. [europepmc]
• End-to-end deep attention-based multitask pipeline for predicting uncertainty-quantified peptide properties from mass spectrometry data. [europepmc]
• De Novo Structure Prediction from Tandem Mass Spectra: Algorithms, Benchmarks, and Limitations. [europepmc]
• Persistence Assessment of Chemicals: Trajectories toward New Approach Methodologies (P-NAMs). [europepmc]
• Decoding the Oxylipin Chemical Space Using Ion Identity Molecular Networking. [europepmc]
• Structure-informed deep generation enables de novo metabolite annotation in untargeted metabolomics. [europepmc]
• Multimodal foundation models in colorectal cancer: from prediction to trustworthy clinical insight. [europepmc]
• Data-Driven Filter for Detector Oscillation Artifacts in Time-of-Flight Mass Spectrometry. [europepmc]
• Peak2Patch: High-Fidelity Functional Group Identification through Attention-Based Fusion of Infrared and Mass Spectra. [europepmc]
• Visible-Near-Infrared Hyperspectral Imaging Enables Nondestructive Identification of Bean Accessions via 1D Spectral Reflectance Analysis. [europepmc]
• AnnoMe: user-defined classification of HR-MS/MS spectra for natural product discovery. [europepmc]
• Advancing the Discovery of Emerging Contaminants: A Leap in Technology and Data. [europepmc]
• Combating Antibacterial Resistance: The Integrative Role of Artificial Intelligence in Bio-Based Product Development. [europepmc]
• JESTR: Joint Embedding Space Technique for Ranking candidate molecules for the annotation of untargeted metabolomics data. [europepmc]
• Targeted Isolation of Prenylated Flavonoids from Paulownia tomentosa Fruit Extracts via AI-Guided Workflow Integrating LC-UV-HRMS/MS. [europepmc]

Discussions

• DreaMS come true ! Just published by the @pluskal-lab.org group at @iocbprague.bsky.social : Self-supervised learning of molecular representations from millions of tandem mass spectra using DreaMS. Pa [bsky, 25 points, 0 comments]
• DreaMS is a neural network able of learning molecular representations from raw unannotated mass spectra through self-supervised learning. An Atlas of 201 millions of MS/Ms data to unravel the unknown [bsky, 5 points, 0 comments]
• Original article: Bushuiev, R.; Bushuiev, A.; Samusevich, R.; Brungs, C.; Sivic, J.; Pluskal, T. Self-supervised learning of molecular representations from millions of tandem mass spectra using DreaMS [bsky, 1 points, 0 comments]

Related

• Efficiently predicting high resolution mass spectra with graph neural networks
• MassSpecGym: A benchmark for the discovery and identification of molecules
• GLACIER: Rethinking Mass Spectrum Prediction as an Object Detection Problem
• Generating Molecular Fragmentation Graphs with Autoregressive Neural Networks
• To Bin or not to Bin: Alternative Representations of Mass Spectra