AI platform ContactSeek cuts gene-editor off-target edits in Nature study

Scientists have described an AI-guided platform that redesigns gene editors for higher precision, according to research covered by phys.org and published in Nature. The China-based team's ContactSeek system uses AlphaFold3 to predict how strongly biological molecules are likely to interact. phys.org reported that AlphaFold3 estimates the probability that amino acids in a gene-editing protein will come into close contact with target DNA and guide RNA. By comparing those contact probabilities for intended and unintended DNA targets, the platform identifies parts of an editor that contribute to accidental edits, differing from approaches that focus on 3D shapes alone. The study authors tested ContactSeek on an adenine base editor. When suggested corrections were applied, the upgraded version reduced genome-wide off-target editing by about 83% for one of the guide RNAs tested, phys.org said. The platform also helped improve a Cas12a-based cytosine base editor. ContactSeek is not yet ready for clinical use because further tests in different cell types and genome targets are needed. The researchers said they anticipate the approach will be broadly applicable to a wide range of DNA- and RNA-editing tools. Hoi Yee Chu and Alan S.L. Wong of the University of Hong Kong published a News and Views piece in Nature on the work. The paper's DOI is 10.1038/s41586-026-10794-z.