Particle.news
Download on the App Store

AI Designs Synthetic CRISPR-Like Enzymes That Work in Bacteria, Plants and Human Cells

The study shows a structure-guided AI workflow can create non-natural RNA-guided nucleases and prompts further safety and specificity testing

Overview

  • Researchers used a hybrid design that combined the ESM-IF1 inverse-folding model with evolution-informed residue constraints to generate new variants of the compact TnpB nuclease.
  • The team screened the AI-designed proteins, called SynTnpBs, first in bacteria and then tested top candidates in plant and human cells to assess editing activity.
  • The paper, which published on Thursday, July 16, 2026, reports many SynTnpBs matched or exceeded the activity of natural TnpB enzymes across those cell types.
  • Cryo-electron microscopy of the most sequence-divergent SynTnpBs revealed new electrostatic and hydrogen-bond networks at the RNA–DNA interface that stabilize active conformations.
  • Next steps include systematic specificity and safety profiling, broader functional tests for research and agricultural use, and consideration of regulatory pathways for any clinical or commercial applications.