Protein and genomic language models uncover the unexplored diversity of bacterial immunity - PubMed
3 hours ago
- #bacterial immunity
- #machine learning
- #genomics
- Machine learning models predict antiphage systems with high precision and recall.
- Experimental validation led to discovery of 12 new antiphage systems.
- Analysis of >32,000 bacterial genomes shows ~1.5% devoted to defense.
- Over 85% of predicted protein families in bacterial immunity are uncharacterized.
- Interactive catalog of >19,000 candidate operon families provided for future study.