Researchers from the University of Zurich, Ghent University, and ETH Zurich have created a new method to edit DNA with high accuracy. This method combines genetic engineering with artificial intelligence (AI). The goal is to make exact changes to DNA without unwanted errors, which could help model human diseases and develop gene therapies, treatments that fix faulty genes.
This research is published in Nature Biotechnology. The method uses CRISPR/CAS9, a gene-editing tool that acts like scissors to cut DNA at specific spots. Normally, when CRISPR cuts DNA, cells repair the break using natural processes, but these can cause mistakes like extra changes or damage to nearby genes. To fix this, the researchers made Pythia, an AI tool named after an ancient oracle that predicts the future. Pythia learns patterns in how cells repair DNA breaks, which are not random but follow rules.
Using AI, Pythia simulates millions of possible repair outcomes and designs small DNA repair templates. These templates are like guides that stick to the cut DNA and direct the cell to make the exact change wanted, such as fixing a single letter in the genetic code or adding a whole gene.
Testing and applications
The method was tested first in human cell cultures, where it allowed precise edits. It also worked in other living things, like xenopus frogs used in research and mice, where brain cells were edited successfully. Besides small changes, it can add fluorescent labels to proteins, glowing markers that let scientists watch how proteins work in healthy or sick tissues.
A key benefit is that it works in all cell types, even those that do not divide, like in the brain. This makes it useful for neurological diseases. Pythia predicts how cells will react to edits, making gene editing safer and more reliable. The researchers explain that ai uses dna repair rules to find the best way to make specific genome changes. They are persuaded that linking AI predictions with real biology opens doors for better gene therapies that are effective and reduce risks.