OneProt: Towards multi-modal protein foundation models via latent space alignment of sequence, structure, binding sites and text encoders
PLOS Computational Biology · 13.11.2025 · 21 (11) · e1013679
DOI: 10.1371/journal.pcbi.1013679 ↗Kurz gefasst
Fachartikel in der Zeitschrift PLOS Computational Biology, online erschienen am 13.11.2025. Verfasst von Klemens Flöge und 10 weiteren Autor:innen. Beteiligte Einrichtung: PriorLabs, Berlin, Germany. Hauptthemen laut MEDLINE: Proteins, Sequence Alignment. Die Arbeit wurde bisher 14-mal zitiert (Crossref) und nennt 64 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Helmholtz Association Initiative and Networking, John von Neumann Institute for Computing (NIC), Branco Weiss Fellowship.
Abstract (Originalsprache)
Recent advances in Artificial Intelligence have enabled multi-modal systems to model and translate diverse information spaces. Extending beyond text and vision, we introduce OneProt, a multi-modal Deep Learning model for proteins that integrates structural, sequence, text, and binding site data. Using the ImageBind framework, OneProt aligns the latent spaces of protein modality encoders in a lightweight fine-tuning scheme that focuses on pairwise alignment with sequence data, rather than requiring full matches. This novel approach comprises a mix of Graph Neural Networks and transformer architectures. It demonstrates good performance in retrieval tasks and showcases the efficacy of multi-modal systems in Protein Machine Learning through a broad spectrum of downstream baselines, including enzyme function prediction and binding site analysis. Furthermore, OneProt enables the transfer of representational information from specialized encoders to the sequence encoder, enhancing capabilities for distinguishing evolutionarily related and unrelated sequences and exhibiting representational properties where evolutionarily related proteins align in similar directions within the latent space. In addition, we extensively investigate modality ablations to identify the encoders that contribute the most to predictive performance, highlighting the significance of the binding site encoder, which has not been used in similar models previously. This work expands the horizons of multi-modal protein models, paving the way for transformative applications in drug discovery, biocatalytic reaction planning, and protein engineering.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Substanzen
Autor:innen und Institutionen
- PriorLabs, Berlin, Germany.
- Helmholtz AI, Munich, Germany.
- Independent Researcher, Coimbatore, Tamil Nadu, India.
- School of Computation, Information and Technology, Technical University of Munich, Garching, Germany.
- Munich Data Science Institute, Technical University of Munich, Garching, Germany.
- Helmholtz Munich, Neuherberg, Germany.
ORCID-Profile: Holger Gohlke, Alina Bazarova
Förderung
- Helmholtz Association Initiative and Networking
- John von Neumann Institute for Computing (NIC)
- Branco Weiss Fellowship
- Helmholtz Association Initiative and Networking Fund
- Helmholtz Foundational Model Initiative
- Gauss Centre for Supercomputing
- Branco Weiss Fellowship – Society in Science
Zitieren
APA
Flöge, K., Udayakumar, S., Sommer, J., Piraud, M., Kesselheim, S., Fortuin, V., Günnemann, S., van der Weg, K. J., Gohlke, H., Merdivan, E., & Bazarova, A. (2025). OneProt: Towards multi-modal protein foundation models via latent space alignment of sequence, structure, binding sites and text encoders. PLOS Computational Biology, 21(11), e1013679. https://doi.org/10.1371/journal.pcbi.1013679
Vancouver
Flöge K, Udayakumar S, Sommer J, Piraud M, Kesselheim S, Fortuin V, et al. OneProt: Towards multi-modal protein foundation models via latent space alignment of sequence, structure, binding sites and text encoders. PLOS Computational Biology. 2025;21(11):e1013679. doi:10.1371/journal.pcbi.1013679
BibTeX
@article{Flge2025,
author = {Flöge, Klemens and Udayakumar, Srisruthi and Sommer, Johanna and Piraud, Marie and Kesselheim, Stefan and Fortuin, Vincent and Günnemann, Stephan and van der Weg, Karel J and Gohlke, Holger and Merdivan, Erinc and Bazarova, Alina},
title = {OneProt: Towards multi-modal protein foundation models via latent space alignment of sequence, structure, binding sites and text encoders},
journal = {PLOS Computational Biology},
year = {2025},
volume = {21},
number = {11},
pages = {e1013679},
doi = {10.1371/journal.pcbi.1013679}
}
EUVIMED