EUVIMEDEUVIMED entdecken
Open Access

Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants

Moritz Ertelt iD, Rocco Moretti iD, Jens Meiler iD, Clara T Schoeder iD

Science Advances · 14.02.2025 · 11 (7)

DOI: 10.1126/sciadv.adr7338 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Science Advances, erschienen am 14.02.2025. Verfasst von Moritz Ertelt und 3 weiteren Autor:innen. Beteiligte Einrichtung: Institute for Drug Discovery, Leipzig University Faculty of Medicine, Leipzig, Germany. Hauptthemen laut MEDLINE: Proteins, Supervised Machine Learning, Machine Learning, Computational Biology. Die Arbeit wurde bisher 19-mal zitiert (Crossref) und nennt 63 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine learning (ML) is changing the world of computational protein design, with data-driven methods surpassing biophysical-based methods in experimental success. However, they are most often reported as case studies, lack integration and standardization, and are therefore hard to objectively compare. In this study, we established a streamlined and diverse toolbox for methods that predict amino acid probabilities inside the Rosetta software framework that allows for the side-by-side comparison of these models. Subsequently, existing protein fitness landscapes were used to benchmark novel ML methods in realistic protein design settings. We focused on the traditional problems of protein design: sampling and scoring. A major finding of our study is that ML approaches are better at purging the sampling space from deleterious mutations. Nevertheless, scoring resulting mutations without model fine-tuning showed no clear improvement over scoring with Rosetta. We conclude that ML now complements, rather than replaces, biophysical methods in protein design.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

ProteinsSupervised Machine LearningMachine LearningComputational BiologyProtein EngineeringSoftwareMutationModels, MolecularAlgorithms

Substanzen

Proteins

Autor:innen und Institutionen

  • Institute for Drug Discovery, Leipzig University Faculty of Medicine, Leipzig, Germany.
  • Center for Scalable Data Analytics and Artificial Intelligence ScaDS.AI, Dresden/Leipzig, Dresden, Germany.
  • Department of Chemistry, Vanderbilt University, Nashville, TN, USA.
  • Center for Structural Biology, Vanderbilt University, Nashville, TN, USA.

ORCID-Profile: Moritz Ertelt, Rocco Moretti, Jens Meiler, Clara T Schoeder

Zitieren

APA

Ertelt, M., Moretti, R., Meiler, J., & Schoeder, C. T. (2025). Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants. Science Advances, 11(7). https://doi.org/10.1126/sciadv.adr7338

Vancouver

Ertelt M, Moretti R, Meiler J, Schoeder CT. Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants. Science Advances. 2025;11(7). doi:10.1126/sciadv.adr7338

BibTeX
@article{Ertelt2025,
  author = {Ertelt, Moritz and Moretti, Rocco and Meiler, Jens and Schoeder, Clara T},
  title = {Self-supervised machine learning methods for protein design improve sampling but not the identification of high-fitness variants},
  journal = {Science Advances},
  year = {2025},
  volume = {11},
  number = {7},
  doi = {10.1126/sciadv.adr7338}
}
Kostenlos starten →