Simplifying in silico protein evolution with minimal screening by unZipro.

unZipro achieves a 61% success rate in identifying high-activity protein variants from minimal libraries across nine diverse proteins.

• Why it matters: Current AI-driven protein engineering methods face challenges like high computational costs and limited success, especially for plant proteins, creating a need for more efficient approaches.
• What they did: The authors developed unZipro, a framework combining a pre-trained inverse folding model with meta-learning to derive family-specific fitness landscapes, tested on approximately 10 candidates per protein.
• The result: unZipro enables cost-effective, zero-shot protein evolution, leading to the creation of superior plant gene-editing nucleases, luciferase variants, rice transcription factors, and barley antiviral proteins with increased potency.