Researchers at the Yong Loo Lin School of Medicine at the National University of Singapore have developed an artificial intelligence-guided method to enhance compact gene-editing tools, improving precision while reducing safety risks in DNA correction.
The study on gene-editing tools demonstrates a strategy to engineer smaller base editors that could advance gene therapy applications by enabling safer and more efficient genetic modifications.
The research, published in Advanced Science, was led by Assistant Professor Jungjoon K. Lee from the Department of Biochemistry and associated synthetic biology programs at NUS Medicine.
The team combined AI-driven protein modeling with a bacterial evolution platform to optimize SsdAtox, a compact DNA-editing enzyme. The engineered variants matched or outperformed leading base editors such as BE4max across multiple human gene targets, while significantly lowering unintended DNA damage and cellular toxicity.
Base editors are designed to correct single-letter mutations in DNA without cutting the genetic strand, offering a safer alternative to earlier gene-editing technologies. However, widely used high-performance editors are large in size, making delivery into cells challenging, particularly when using adeno-associated viruses commonly employed in gene therapy. Increasing editing efficiency has also been associated with higher risks of unintended genetic alterations.
SsdAtox, which is smaller than conventional base editors, has shown potential for therapeutic delivery but suffers from low efficiency and higher toxicity in its natural form. The research focused on improving its performance while maintaining its compact structure. The team applied AlphaFold3, an AI system for predicting protein structures, to identify and modify a key region controlling DNA access to the enzyme’s active site, enhancing its activity.
To further refine the enzyme, researchers developed a screening platform called Trinity-Screen. This system simultaneously evaluates editing efficiency, DNA safety, and cellular toxicity within bacterial cells. Only variants meeting all criteria progressed through multiple selection stages before being tested in human cells across artificial and natural gene targets.
Optimized, more safe gene-editing variants
The optimized variants demonstrated up to 11.8-fold higher editing efficiency, reduced DNA break rates by nearly half compared to earlier high-activity versions, and showed significantly lower toxicity. One leading variant reduced unwanted DNA breaks by 37 percent compared to BE4max, while maintaining a more consistent editing range.
The team also introduced a new evaluation metric, the Base Editor Performance Index, which balances efficiency and safety to provide a comprehensive assessment of gene-editing tools. The smaller size of the optimized enzymes supports easier delivery into viral systems, potentially expanding treatment options for genetic diseases.
The study highlights the application of AI-guided protein modeling in directly improving genome-editing enzymes and offers a scalable framework for developing safer gene-editing technologies.
Trending | GACA mobilizes full capacity to boost regional air traffic
