Benchmarking Generative AI Protein Models Reveals Differences Between Structural and Sequence-based Approaches
Genomics, Proteomics & Bioinformatics · 26.06.2026 · 24 (2)
DOI: 10.1093/gpbjnl/qzag014 ↗Kurz gefasst
Fachartikel in der Zeitschrift Genomics, Proteomics & Bioinformatics, online erschienen am 15.02.2026. Verfasst von Alexander J Barnett und 6 weiteren Autor:innen. Beteiligte Einrichtung: Menzies Institute for Medical Research, University of Tasmania, Hobart, TAS 7000, Australia. Hauptthemen laut MEDLINE: Models, Molecular, Proteins. Die Arbeit wurde bisher 3-mal zitiert (Crossref) und nennt 39 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Recent advances in artificial intelligence have led to the development of generative models for de novo protein design. In this study, we compared 13 state-of-the-art generative protein models, assessing their ability to produce feasible, diverse, and novel protein monomers. Structural diffusion models generally create designs with higher confidence in predicted structures and more biologically plausible energy distributions, but exhibit limited diversity and strong sequence biases. Conversely, protein language models generate more diverse and novel designs but with lower structural confidence. We also evaluated the ability of these models to generate unique proteins, conditionally based on the tobacco etch virus (TEV) protease. Generative models are successful in producing functional enzymes, albeit with diminished activity compared to the wild-type TEV. Our systematic benchmarking provides a foundation for evaluating and selecting generative protein models, while highlighting the complementary strengths of different generative paradigms. This framework will facilitate informed application of these tools for biomedical engineering and design.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Substanzen
Autor:innen und Institutionen
- Menzies Institute for Medical Research, University of Tasmania, Hobart, TAS 7000, Australia.
- Alphinia Bio, Hobart, TAS 7008, Australia.
- School of Medicine, University of Tasmania, Hobart, TAS 7000, Australia.
- Centre for Eye Research Australia, University of Melbourne, Melbourne, VIC 3002, Australia.
ORCID-Profile: Alexander J Barnett, Rajendra Kc, Pratikshya Pandey, Pamodha Somasiri, Kirsten A Fairfax, Sandy Hung, Alex W Hewitt
Zitieren
APA
Barnett, A. J., Kc, R., Pandey, P., Somasiri, P., Fairfax, K. A., Hung, S., & Hewitt, A. W. (2026). Benchmarking Generative AI Protein Models Reveals Differences Between Structural and Sequence-based Approaches. Genomics, Proteomics & Bioinformatics, 24(2). https://doi.org/10.1093/gpbjnl/qzag014
Vancouver
Barnett AJ, Kc R, Pandey P, Somasiri P, Fairfax KA, Hung S, et al. Benchmarking Generative AI Protein Models Reveals Differences Between Structural and Sequence-based Approaches. Genomics, Proteomics & Bioinformatics. 2026;24(2). doi:10.1093/gpbjnl/qzag014
BibTeX
@article{Barnett2026,
author = {Barnett, Alexander J and Kc, Rajendra and Pandey, Pratikshya and Somasiri, Pamodha and Fairfax, Kirsten A and Hung, Sandy and Hewitt, Alex W},
title = {Benchmarking Generative AI Protein Models Reveals Differences Between Structural and Sequence-based Approaches},
journal = {Genomics, Proteomics & Bioinformatics},
year = {2026},
volume = {24},
number = {2},
doi = {10.1093/gpbjnl/qzag014}
}
EUVIMED