EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Protein language models learn evolutionary statistics of interacting sequence motifs

Zhidian Zhang, Hannah K Wayment-Steele, Garyk Brixi, Haobo Wang, Dorothee Kern iD, Sergey Ovchinnikov iD

Proceedings of the National Academy of Sciences · 05.11.2024 · 121 (45)

DOI: 10.1073/pnas.2406285121 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the National Academy of Sciences, online erschienen am 28.10.2024. Verfasst von Zhidian Zhang und 5 weiteren Autor:innen. Beteiligte Einrichtung: Harvard University, Cambridge, MA 02138. Hauptthemen laut MEDLINE: Proteins. Die Arbeit wurde bisher 159-mal zitiert (Crossref) und nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Institutes of Health (NIH), Howard Hughes Medical Institute (HHMI), Swiss National Science Foundation (SNSF).

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Protein language models (pLMs) have emerged as potent tools for predicting and designing protein structure and function, and the degree to which these models fundamentally understand the inherent biophysics of protein structure stands as an open question. Motivated by a finding that pLM-based structure predictors erroneously predict nonphysical structures for protein isoforms, we investigated the nature of sequence context needed for contact predictions in the pLM Evolutionary Scale Modeling (ESM-2). We demonstrate by use of a “categorical Jacobian” calculation that ESM-2 stores statistics of coevolving residues, analogously to simpler modeling approaches like Markov Random Fields and Multivariate Gaussian models. We further investigated how ESM-2 “stores” information needed to predict contacts by comparing sequence masking strategies, and found that providing local windows of sequence information allowed ESM-2 to best recover predicted contacts. This suggests that pLMs predict contacts by storing motifs of pairwise contacts. Our investigation highlights the limitations of current pLMs and underscores the importance of understanding the underlying mechanisms of these models.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

ProteinsEvolution, MolecularAmino Acid MotifsModels, MolecularProtein ConformationMarkov Chains

Stichwörter der Autor:innen

interpretability studylanguage modelsprotein structure prediction

Substanzen

Proteins

Autor:innen und Institutionen

  • Harvard University, Cambridge, MA 02138.
  • Department of Biology, Massachusetts Institute of Technology, Cambridge, MA 02139.
  • Institute of Bioengineering, School of Life Sciences, Ecole polytechnique fédérale de Lausanne, Lausanne VD 1015, Switzerland.
  • HHMI, Brandeis University, Waltham, MA 02453.
  • Department of Biochemistry, Brandeis University, Waltham, MA 02453.
  • Harvard College, Harvard University, Cambridge, MA 02138.

ORCID-Profile: Dorothee Kern, Sergey Ovchinnikov

Förderung

  • National Institutes of Health (NIH)
  • Howard Hughes Medical Institute (HHMI)
  • Swiss National Science Foundation (SNSF)
  • National Foundation of Science (NSF)
  • NIH HHS
  • Jane Coffin Childs

Zitieren

APA

Zhang, Z., Wayment-Steele, H. K., Brixi, G., Wang, H., Kern, D., & Ovchinnikov, S. (2024). Protein language models learn evolutionary statistics of interacting sequence motifs. Proceedings of the National Academy of Sciences, 121(45). https://doi.org/10.1073/pnas.2406285121

Vancouver

Zhang Z, Wayment-Steele HK, Brixi G, Wang H, Kern D, Ovchinnikov S. Protein language models learn evolutionary statistics of interacting sequence motifs. Proceedings of the National Academy of Sciences. 2024;121(45). doi:10.1073/pnas.2406285121

BibTeX
@article{Zhang2024,
  author = {Zhang, Zhidian and Wayment-Steele, Hannah K and Brixi, Garyk and Wang, Haobo and Kern, Dorothee and Ovchinnikov, Sergey},
  title = {Protein language models learn evolutionary statistics of interacting sequence motifs},
  journal = {Proceedings of the National Academy of Sciences},
  year = {2024},
  volume = {121},
  number = {45},
  doi = {10.1073/pnas.2406285121}
}
Kostenlos starten →