Separating selection from mutation in antibody language models
eLife · 07.04.2026 · 15
DOI: 10.7554/elife.109644 ↗Kurz gefasst
Fachartikel in der Zeitschrift eLife, online erschienen am 07.04.2026. Verfasst von Frederick A Matsen und 8 weiteren Autor:innen. Beteiligte Einrichtung: Computational Biology Program, Fred Hutchinson Cancer Center, Seattle, United States. Hauptthemen laut MEDLINE: Antibodies, Mutation, Selection, Genetic. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 70 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: NIAID NIH HHS, NIH HHS, NHGRI NIH HHS.
Abstract (Originalsprache)
Antibodies are encoded by nucleotide sequences that are generated by V(D)J recombination and evolve according to mutation and selection processes. Existing antibody language models, however, focus exclusively on antibodies as strings of amino acids and are fitted using standard language modeling objectives such as masked or autoregressive prediction. In this paper, we first show that fitting models using this objective implicitly incorporates nucleotide-level mutation processes as part of the protein language model, which degrades performance when predicting effects of mutations on functional properties of antibodies. To address this limitation, we devise a new framework: a deep amino acid selection model (DASM) that learns the selection effects of amino acid mutations while explicitly factoring out the nucleotide-level mutation process. By fitting selection as a separate term from the mutation process, the DASM exclusively quantifies functional effects: effects that change some aspect of the function of the antibody. This factorization leads to substantially improved performance on standard functional benchmarks. Moreover, our model is an order of magnitude smaller and multiple orders of magnitude faster to evaluate than existing approaches, as well as being readily interpretable.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Computational Biology Program, Fred Hutchinson Cancer Center, Seattle, United States.
- Department of Genome Sciences, University of Washington, Seattle, United States.
- Department of Statistics, University of Washington, Seattle, United States.
- Howard Hughes Medical Institute, Seattle, United States.
- Department of Biochemistry, University of Utah School of Medicine, Salt Lake City, United States.
- Computer Science Division and Department of Statistics, University of California, Berkeley, Berkeley, United States.
ORCID-Profile: Frederick A Matsen, Will Dumm, Kevin Sung, Mackenzie M Johnson, Tyler N Starr, Yun S Song, Julia Fukuyama, Hugh K Haddox
Förderung
- NIAID NIH HHS
- NIH HHS
- NHGRI NIH HHS
- National Institutes of Health
- Searle Scholars Program
- Howard Hughes Medical Institute
Zitieren
APA
Matsen, F. A., Dumm, W., Sung, K., Johnson, M. M., Rich, D. H., Starr, T. N., Song, Y. S., Fukuyama, J., & Haddox, H. K. (2026). Separating selection from mutation in antibody language models. eLife, 15. https://doi.org/10.7554/elife.109644
Vancouver
Matsen FA, Dumm W, Sung K, Johnson MM, Rich DH, Starr TN, et al. Separating selection from mutation in antibody language models. eLife. 2026;15. doi:10.7554/elife.109644
BibTeX
@article{Matsen2026,
author = {Matsen, Frederick A and Dumm, Will and Sung, Kevin and Johnson, Mackenzie M and Rich, David H and Starr, Tyler N and Song, Yun S and Fukuyama, Julia and Haddox, Hugh K},
title = {Separating selection from mutation in antibody language models},
journal = {eLife},
year = {2026},
volume = {15},
doi = {10.7554/elife.109644}
}
EUVIMED