Deep learning boosts precision of CRISPR genome editing
Prof. Soeren Lienkamp's group has developed a new strategy to make CRISPR-based genome editing more accurate and predictable. Using deep-learning models, they designed short, repeated DNA sequence, microhomology tandem repeats, that guide the cell’s natural DNA repair machinery to precisely integrate new genetic material without causing unwanted deletions.
The method enables targeted insertion of large DNA fragments, precise protein tagging, and accurate small DNA changes om frogs and mice. It works in cell types where traditional repair pathways are inefficient and is supported by a new design tool called Pythia. This approach opens the door to more reliable genetic engineering for both research and potential therapeutic applications.
The study appears in Nature Biotechnology.