Direct discovery · Biology · Medicine
OpenCRISPR-1 shows AI-generated genome editing in human cells
AI-generated Cas9-like genome editors, including OpenCRISPR-1, experimentally tested in human cells.
Summary
Researchers used protein language models trained on CRISPR-Cas diversity to generate new Cas9-like editors and guide-RNA components. In human-cell experiments, multiple generated nucleases showed editing activity, and the nominated OpenCRISPR-1 editor had comparable on-target activity with reduced off-target editing in the reported test panel.
AI role
Protein language models generated CRISPR-Cas protein and guide-RNA designs after mining large microbial and metagenomic sequence datasets.
Narrative role
This is a concrete wet-lab example of AI generating functional genome-editing proteins rather than only ranking natural candidates or predicting structures.
Caveat
The result is a research-tool demonstration in controlled assays; therapeutic usefulness, broad target performance, delivery, and safety require further validation.