EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

EvoSeq-ML: advancing data-centric machine learning with evolutionary-informed protein sequence representation and generation

Mehrsa Mardikoraem iD, Nathaniel Pascual iD, Joelle N Eaves, Shramana Chatterjee, Logan Garland, James VanAntwerp, Anton Mahama, Patrick Finneran, Robert P Hausinger iD, Daniel R Woldring iD

Protein Engineering, Design and Selection · 09.01.2026 · 39

DOI: 10.1093/protein/gzag033 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Protein Engineering, Design and Selection, online erschienen am 03.11.2026. Verfasst von Mehrsa Mardikoraem und 9 weiteren Autor:innen. Beteiligte Einrichtung: Department of Chemical Engineering and Materials Science, Michigan State University, 428 S. Shaw Lane, East Lansing, MI 48824, United States. Hauptthemen laut MEDLINE: Protein Engineering, Machine Learning, Evolution, Molecular, Proteins. Die Arbeit nennt 62 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: USDA, National Science Foundation.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

From protein structure prediction to novel protein generation, challenging protein engineering tasks have been made possible by advancements in machine learning (ML). While largely driven by ML architecture refinements, these advancements in ML-based protein engineering campaigns have left the impact of data curation underexplored. In light of the growing wealth of labeled sequence data, data-centric advances (e.g. prioritizing improvements in ML protein engineering tools through the curation of high-quality, domain-specific training data) are increasingly preferred over model-centric advancements. Implementing datasets that accurately reflect biological complexity and diversity has been shown to improve the efficiency of training protein engineering ML tools. Here, we evaluate an ancestral sequence reconstruction (ASR)-informed data augmentation strategy for training generative and representation-learning models in protein engineering. Using ethylene-forming enzyme (EFE) as a model system, we show that variational autoencoder models trained on ancestral and near-ancestral sequence datasets generate variants with improved predicted and experimentally measured thermostability relative to variants generated from modern-sequence training data. All experimentally tested ancestral and ML-generated EFEs produced detectable ethylene, although ML-generated variants showed reduced activity relative to wild-type EFE, indicating that the approach more strongly captured stability-associated features than catalytic optimization. We further evaluated ASR-enriched sequence sets for fine-tuning ESM2 representations in endolysin and lysozyme C stability-classification tasks, where ancestral representations were competitive with modern-sequence fine-tuning in selected settings. Overall, this work supports ASR-informed data augmentation as a promising strategy for stability-oriented protein sequence generation and motivates future work to couple ancestral sequence diversity with explicit functional selection.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

Protein EngineeringMachine LearningEvolution, MolecularProteinsAmino Acid SequenceAutoencoder

Stichwörter der Autor:innen

Ancestral Sequence ReconstructionData-Centric ModelsFine-TuningGenerative ModelsLanguage ModelsProtein Engineering

Substanzen

Proteins

Autor:innen und Institutionen

  • Department of Chemical Engineering and Materials Science, Michigan State University, 428 S. Shaw Lane, East Lansing, MI 48824, United States.
  • Institute for Quantitative Health Science and Engineering, Michigan State University, 775 Woodlot Drive, East Lansing, MI 48824, United States.
  • Department of Microbiology, Genetics, and Immunology, Michigan State University, 567 Wilson Rd, East Lansing, MI 48824, United States.
  • Department of Chemical & Biomolecular Engineering, University of Delaware, 150 Academic Steet, Newark, DE 19716, United States.
  • Department of Molecular Pharmacology, School of Human Medicine, University of Pittsburgh, M240 Alan Magee Scaife Hall, Pittsburgh, PA 15261, United States.
  • Biophysics, Menten AI, 528 41st Avenue, San Francisco, CA 94121, United States.

ORCID-Profile: Mehrsa Mardikoraem, Nathaniel Pascual, Robert P Hausinger, Daniel R Woldring

Förderung

  • USDA
  • National Science Foundation

Zitieren

APA

Mardikoraem, M., Pascual, N., Eaves, J. N., Chatterjee, S., Garland, L., VanAntwerp, J., Mahama, A., Finneran, P., Hausinger, R. P., & Woldring, D. R. (2026). EvoSeq-ML: advancing data-centric machine learning with evolutionary-informed protein sequence representation and generation. Protein Engineering, Design and Selection, 39. https://doi.org/10.1093/protein/gzag033

Vancouver

Mardikoraem M, Pascual N, Eaves JN, Chatterjee S, Garland L, VanAntwerp J, et al. EvoSeq-ML: advancing data-centric machine learning with evolutionary-informed protein sequence representation and generation. Protein Engineering, Design and Selection. 2026;39. doi:10.1093/protein/gzag033

BibTeX
@article{Mardikoraem2026,
  author = {Mardikoraem, Mehrsa and Pascual, Nathaniel and Eaves, Joelle N and Chatterjee, Shramana and Garland, Logan and VanAntwerp, James and Mahama, Anton and Finneran, Patrick and Hausinger, Robert P and Woldring, Daniel R},
  title = {EvoSeq-ML: advancing data-centric machine learning with evolutionary-informed protein sequence representation and generation},
  journal = {Protein Engineering, Design and Selection},
  year = {2026},
  volume = {39},
  doi = {10.1093/protein/gzag033}
}
Kostenlos starten →