EUVIMEDEUVIMED entdecken
Open Access

SeaMoon: From protein language models to continuous structural heterogeneity

Valentin Lombard, Dan Timsit, Sergei Grudinin, Elodie Laine

Structure · 2025-09 · 33 (9) · 1577-1590.e8

DOI: 10.1016/j.str.2025.06.010 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Structure, 2025-09. Verfasst von Valentin Lombard und 3 weiteren Autor:innen. Beteiligte Einrichtung: Sorbonne Université, CNRS, IBPS, Department of Computational, Quantitative and Synthetic Biology (CQSB, UMR7238), 75005 Paris, France. Hauptthemen laut MEDLINE: Proteins, Software. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 111 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: French National Research Agency, European Research Council, European Commission.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

How proteins move and deform determines their interactions with the environment and is thus of the utmost importance for cellular functioning. Following the revolution in single protein 3D structure prediction, researchers have focused on repurposing or developing deep learning models for sampling alternative protein conformations. In this work, we explored whether continuous compact representations of protein motions could be predicted directly from sequences, without exploiting 3D structures. SeaMoon leverages protein language model (pLM) embeddings as input to a lightweight convolutional neural network. We assessed SeaMoon against ∼1,000 collections of experimental conformations exhibiting diverse motions. It predicts at least one ground-truth motion with reasonable accuracy for 40% of the test proteins. SeaMoon captures motions inaccessible to normal mode analysis, an unsupervised physics-based method relying solely on 3D geometry, and generalizes to proteins without detectable sequence similarity to the training set. SeaMoon is easily retrainable with novel or updated pLMs.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

ProteinsSoftwareProtein ConformationModels, MolecularNeural Networks, ComputerDeep Learning

Stichwörter der Autor:innen

PCAconvolutional neural networkdeep learningprincipal component predictionprotein language modelsprotein motion predictionsubspace predictiontransfer learning

Substanzen

Proteins

Autor:innen und Institutionen

  • Sorbonne Université, CNRS, IBPS, Department of Computational, Quantitative and Synthetic Biology (CQSB, UMR7238), 75005 Paris, France.
  • Université Grenoble Alpes, CNRS, Grenoble INP, LJK, 38000 Grenoble, France. Electronic address: sergei.grudinin@univ-grenoble-alpes.fr.
  • Sorbonne Université, CNRS, IBPS, Department of Computational, Quantitative and Synthetic Biology (CQSB, UMR7238), 75005 Paris, France; Institut Universitaire de France (IUF), Paris, France. Electronic address: elodie.la…

Förderung

  • French National Research Agency
  • European Research Council
  • European Commission
  • Sorbonne Université

Zitieren

APA

Lombard, V., Timsit, D., Grudinin, S., & Laine, E. (2025). SeaMoon: From protein language models to continuous structural heterogeneity. Structure, 33(9), 1577-1590.e8. https://doi.org/10.1016/j.str.2025.06.010

Vancouver

Lombard V, Timsit D, Grudinin S, Laine E. SeaMoon: From protein language models to continuous structural heterogeneity. Structure. 2025;33(9):1577-1590.e8. doi:10.1016/j.str.2025.06.010

BibTeX
@article{Lombard2025,
  author = {Lombard, Valentin and Timsit, Dan and Grudinin, Sergei and Laine, Elodie},
  title = {SeaMoon: From protein language models to continuous structural heterogeneity},
  journal = {Structure},
  year = {2025},
  volume = {33},
  number = {9},
  pages = {1577--1590.e8},
  doi = {10.1016/j.str.2025.06.010}
}
Kostenlos starten →