Structure-guided inverse protein folding coupled with evolution-informed residue constraints enables the design of highly divergent RNA-guided TnpB nucleases that achieve 24% experimental success rates and retain or exceed wild-type genome editing activity across bacterial, plant, and human cells through AI-generated residues that create new stabilizing RNA-DNA interface contacts while preserving essential conformational dynamics required for nuclease function.

The design of RNA-guided nucleases with properties not limited by evolution can expand programmable genome editing capabilities. However, generating diverse multi-domain proteins with robust enzymatic properties remains challenging. Here we use an artificial intelligence-driven strategy that couples structure-guided inverse protein folding with evolution-informed residue constraints to generate active, divergent variants of TnpB, a minimal CRISPR-Cas12-like nuclease. High-throughput functional screening of AI-generated variants yielded editors that retained or exceeded wild-type activity in bacterial, plant and human cells. Cryo-EM-based structure determination of the most divergent active variant revealed new stabilizing contacts in the RNA/DNA interfaces across conformational states, demonstrating the design potential of this approach. Together these results establish a strategy for creating non-natural RNA-guided nucleases and conformationally active nucleic acid binders, enlarging the designable protein space.

An evolution- and structure-conditioned model enables design of active RNA-guided nucleases with new nucleic acid contacts resolved by cryo-EM.