Learning protein representations with conformational dynamics
Bioinformatics · 03.05.2026 · 42 (5)
DOI: 10.1093/bioinformatics/btag254 ↗Kurz gefasst
Fachartikel in der Zeitschrift Bioinformatics, online erschienen am 05.05.2026. Verfasst von Dan Kalifa und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Computer Science, Technion-Israel Institute of Technology, Technion City, Haifa 3200003, Israel. Hauptthemen laut MEDLINE: Protein Conformation, Proteins, Machine Learning, Computational Biology. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 28 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Motivation Proteins change shape as they work, and these changing states control whether binding sites are exposed, signals are relayed, and catalysis proceeds. Most protein language models (PLMs) pair a sequence with a single structural snapshot, which can miss state-dependent features central to interaction, localization, and enzyme activity. Studies also indicate that many proteins assume multiple, functionally relevant shapes, motivating approaches that learn from this variability. Results We present DynamicsPLM, a PLM conditioned on ensembles of computationally generated conformations to derive state-aware representations. DynamicsPLM improves predictive performance across protein–protein interaction, subcellular localization, enzyme classification, and metal-ion binding. On a widely used protein–protein interaction benchmark, it achieves a four-point accuracy gain over the strongest baseline. On a curated test set enriched for proteins with multiple conformational states, the margin increases to eleven points. These findings argue for a shift from static to dynamics-aware modeling, in which conformational variability is treated as informative. By elevating conformational state to a central element of machine learning in protein biology, this work advances modeling toward mechanisms that better reflect how proteins operate in cells and provides a route to actionable hypotheses about when and how binding, signaling, and catalysis occur. Availability and implementation Code, model weights, and inference scripts are available at https://github.com/kalifadan/DynamicsPLM (DOI: https://doi.org/10.5281/zenodo.17668302).
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Substanzen
Autor:innen und Institutionen
- Department of Computer Science, Technion-Israel Institute of Technology, Technion City, Haifa 3200003, Israel.
- Adaptive Systems and Interaction Group, Microsoft Research, Redmond, WA 98052, United States.
- Department of Biomedical Informatics and Medical Education, University of Washington, Seattle, WA 98195, United States.
ORCID-Profile: Dan Kalifa, Eric Horvitz, Kira Radinsky
Zitieren
APA
Kalifa, D., Horvitz, E., & Radinsky, K. (2026). Learning protein representations with conformational dynamics. Bioinformatics, 42(5). https://doi.org/10.1093/bioinformatics/btag254
Vancouver
Kalifa D, Horvitz E, Radinsky K. Learning protein representations with conformational dynamics. Bioinformatics. 2026;42(5). doi:10.1093/bioinformatics/btag254
BibTeX
@article{Kalifa2026,
author = {Kalifa, Dan and Horvitz, Eric and Radinsky, Kira},
title = {Learning protein representations with conformational dynamics},
journal = {Bioinformatics},
year = {2026},
volume = {42},
number = {5},
doi = {10.1093/bioinformatics/btag254}
}
EUVIMED