AI makes gene editing more accurate and reduces errors by up to 83%

The research was commented on by Hoi Yee Chu and Alan SL Wong, from the University of Hong Kong, in an article like News and Views published in the same edition of the magazine, in which they highlight the relevance of the results.

Gene editing is a technique that allows scientists to insert, remove, modify or replace DNA bases in living organisms. The technology has applications in different areas, including correcting disease-associated mutations and developing improved agricultural crops.

Tools, such as CRISPR, work like “molecular scissors”, capable of locating specific points in the genome to make these changes.

Despite its accuracy, the method still has limitations. In some cases, genetic editors can modify incorrect sections of DNA or even RNA, generating unintended changes.

AI identifies points responsible for errors

• To address this problem, scientists in China have developed a platform called ContactSeekAI-driven;
• The system uses the program AlphaFold3a tool capable of predicting the intensity with which different biological molecules tend to interact;
• When a gene editor is introduced into a cell, it needs to bind to a specific sequence of DNA. AlphaFold3 estimates the probability that the amino acids of the protein responsible for editing establish contact with both the target DNA and the guide RNA, which directs the tool to the correct location;
• The platform compares these probabilities between the desired targets and those that should not be modified. This way, it can identify which parts of the genetic editor contribute to accidental changes.

According to the researchers, this strategy differs from the conventional approach, which seeks to identify possible errors by analyzing the three-dimensional structures of the editor and DNA molecules.

“ContactSeek represents a computational approach to improving the specificity of genomic editors,” the scientists wrote. “The contact probability is a more sensitive result than the predicted 3D structure for distinguishing differential interactions.”

Testing reduced off-target edits by around 83%

To evaluate the technology, the researchers applied ContactSeek to an adenine base editor, a tool used to correct single-letter errors in DNA. The AI ​​analyzed the points where the editor made unwanted changes and suggested small modifications to its structure.

Continues after advertising

After implementing these changes, the enhanced version of the editor reduced approximately 83% genome-wide off-target edits for one of the evaluated guide RNAs.

Furthermore, the platform has also been used to increase the accuracy of another type of gene editing tool: a cytosine base editor based on the Cas12a enzyme.

Technology still needs further testing

Despite promising results, ContactSeek is not yet ready for clinical application. According to the researchers, the platform needs to undergo new tests involving different types of cells and different genomic targets before it can be used in treatments.

The team’s expectation, however, is that the technology can be used widely in the future. “We anticipate that ContactSeek… will be broadly applicable to a wide range of DNA and RNA editing tools,” the study authors said.