Protein and genomic language models uncover the unexplored diversity of bacterial immunity
with the discovery of 12 antiphage systems. Applied to over 32,000 bacterial genomes, these models expand the predicted antiphage repertoire, with ~1.5% of bacterial genomes devoted to defense and more than 85% of predicted protein families remaining uncharacterized. We provide an interactive catalog of more than 19,000 candidate operon families for experimental follow-up. Together, these findings show that most molecular diversity in bacterial immunity remains uncharacterized and provide a foundation for its systematic exploration.
Authors
- Helena Shomar (ORCID: https://orcid.org/0000-0002-1090-2871)
- Aude Bernheim (ORCID: https://orcid.org/0000-0003-0212-777X)
- Florian Tesson (ORCID: https://orcid.org/0000-0003-4038-1154)
- Rachel Lavenir
- Ernest Mordret (ORCID: https://orcid.org/0000-0001-6641-8336)
- Alexandre Hervé
- Jean Cury (ORCID: https://orcid.org/0000-0002-6462-8783)
- Rémi Planel (ORCID: https://orcid.org/0000-0002-7826-3316)
- Hugo Vaysset (ORCID: https://orcid.org/0009-0007-3134-1966)
- Tyler Clabby
- Arthur Loubat
Institutions
- Délégation Paris 5 (FR)
- Centre National de la Recherche Scientifique (FR)
- Inserm (FR)
- Institut Pasteur (FR)
- Université Paris Cité (FR)
- AgroParisTech (FR)
- Université Paris-Saclay (FR)
- Sorbonne Paris Cité (FR)
Publication Details
- Journal
- Science
- Published
- 2026-04-02
- DOI
- https://doi.org/10.1126/science.adv8275
- Citations
- 1
- Primary Topic
- Machine Learning in Bioinformatics
- Type
- article
- Field-Weighted Citation Impact
- 12.65
Funders
- European Research Council
- MSDAVENIR