Breakthrough Tracker record
AI plus evolutionary constraints produced compact, active RNA-guided genome editors
Researchers combined machine-learning inverse folding with evolutionary and structural constraints to design synthetic TnpB nucleases. Experimentally screened variants matched or exceeded natural TnpB activity in bacterial, plant and human cells, and cryo-EM revealed stabilizing contacts created by the designs.
← Back to the filtered Breakthrough Tracker
- Stable ID
science-2026-ai-designed-tnpb-nucleases- Revision
science-2026-ai-designed-tnpb-nucleases.v1- Field
- Biology · Protein design and genome editing
- Evidence
- Tier 2 · Peer reviewed: Yes
- Record state
- Current · Peer-reviewed laboratory result
- Last checked
AI role
Machine-learning inverse folding was a core sequence-design method, combined with evolutionary and structural constraints.
Record details
- Problem or result
- Many powerful genome editors are difficult to package and deliver, while unconstrained AI protein designs often fail when tested experimentally.
- Authors
- Petr Skopintsev, Isabel Esain-Garcia, Evan C. DeTurk et al.
- Institutions
- Innovative Genomics Institute; University of California, Berkeley; University of California, Los Angeles
- Result date
- July 16, 2026
Why it matters
The study closes a credible loop from AI sequence generation to cell activity and structural explanation while producing editors smaller than commonly used Cas9 proteins.
Limits
This is laboratory and cell evidence, not a therapeutic result. Editing activity does not establish genome-wide specificity, delivery, immunogenicity or safety, and only a small fraction of generated candidates passed experimental screening.
Sources
- Primary: Science primary paper
- Primary: Preprint and methods context
- Primary: Cryo-EM structure record
- Independent: Nature specialist reporting
Correction and revision history
- 2025-12-08 — Preprint posted.
- 2026-07-16 — Peer-reviewed version published in Science.
Machine-readable: JSON v1 · CSV v1 · Schema v1
This individual record remains noindex until a story-specific featured image passes Kingy’s rendered-pixel visual review. The source-linked tracker hub remains the public index.