Evolutionary Constraints Improve Protein Large Language Model Predictions for Protein Stability, Binding Regions, and Epistasis
Computational and Structural Biotechnology Journal · 01.03.2026 · 35 (1)
DOI: 10.34133/csbj.0226 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computational and Structural Biotechnology Journal, online erschienen am 18.09.2026. Verfasst von Konstantina Tzavella und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Medical Physics, Institut Jules Bordet, Hôpital Universitaire de Bruxelles (H.U.B), Université Libre de Bruxelles (ULB), Brussels 1070,…. Die Arbeit nennt 49 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar. Gefördert von: Fonds Wetenschappelijk Onderzoek, Vrije Universiteit Brussel, Horizon 2020 Framework Programme.
Abstract (Originalsprache)
Our understanding of protein function and evolution is largely based on the relationship between amino acid sequence and overall fold, now effectively captured by computational models. Yet, predicting how mutations—shaped by epistasis—alter protein behavior, especially in dynamic or structurally ambiguous regions, remains difficult. Here, we present D2D, which combines a self-supervised protein language model with protein-specific evolutionary information, to predict mutational effects using little to no task-specific labeled data. D2D captures long-range epistatic interactions and accurately predicts single and higher-order mutation effects on protein thermostability and binding, without being trained on the task. When fine-tuned, D2D outperforms state-of-the-art methods on latent driver cancer mutations and co-occurring proliferation-enhancing mutations across independent experimental studies. Unlike most existing approaches, D2D avoids biases linked to solvent accessibility or to multiple sequence alignment depth and quality, making it particularly effective for disordered or surface binding regions where structure-based predictors typically falter. Overall, D2D provides a general framework for modeling mutational effects in proteins with limited experimental or structural information.
Autor:innen und Institutionen
- Department of Medical Physics, Institut Jules Bordet, Hôpital Universitaire de Bruxelles (H.U.B), Université Libre de Bruxelles (ULB), Brussels 1070, Belgium.
- Interuniversity Institute of Bioinformatics (IB2), ULB-VUB, Brussels, Belgium.
- Brussels Interuniversity Genomics High Throughput Core (BRIGHTcore), Vrije Universiteit Brussel (VUB), Université Libre de Bruxelles (ULB), Brussels, Belgium.
- Clinical Sciences, Research Group Reproduction and Genetics, Centre for Medical Genetics, Vrije Universiteit Brussel (VUB), Universitair Ziekenhuis Brussel (UZ Brussel), Brussels 1090, Belgium.
- Structural Biology Brussels, Vrije Universiteit Brussel, Brussels, Belgium.
- Chemistry Department, Vrije Universiteit Brussel, Brussels, Belgium.
ORCID-Profile: Konstantina Tzavella, Catharina Olsen, Wim Vranken
Förderung
- Fonds Wetenschappelijk Onderzoek
- Vrije Universiteit Brussel
- Horizon 2020 Framework Programme
Zitieren
APA
Tzavella, K., Olsen, C., & Vranken, W. (2026). Evolutionary Constraints Improve Protein Large Language Model Predictions for Protein Stability, Binding Regions, and Epistasis. Computational and Structural Biotechnology Journal, 35(1). https://doi.org/10.34133/csbj.0226
Vancouver
Tzavella K, Olsen C, Vranken W. Evolutionary Constraints Improve Protein Large Language Model Predictions for Protein Stability, Binding Regions, and Epistasis. Computational and Structural Biotechnology Journal. 2026;35(1). doi:10.34133/csbj.0226
BibTeX
@article{Tzavella2026,
author = {Tzavella, Konstantina and Olsen, Catharina and Vranken, Wim},
title = {Evolutionary Constraints Improve Protein Large Language Model Predictions for Protein Stability, Binding Regions, and Epistasis},
journal = {Computational and Structural Biotechnology Journal},
year = {2026},
volume = {35},
number = {1},
doi = {10.34133/csbj.0226}
}
EUVIMED