NUS researchers improve compact base editors with AI for safer gene therapy

NUS researchers improve compact base editors with AI for safer gene therapy

Using AI protein modeling and a new screening system, scientists enhance a small DNA-editing enzyme to achieve higher efficiency with less unwanted damage and toxicity.
GP
Giulio Prisco
Mar 19, 2026
2 min read

Researchers at the National University of Singapore have created a new way to make base editors better. Base editors are tools that fix single-letter errors in DNA without cutting the DNA strand completely. This approach can correct mutations that cause many inherited diseases and may be safer than older gene-editing methods that break DNA fully.

Many strong base editors are too large to fit easily into delivery vehicles like adeno-associated viruses (AAVs), which are harmless viruses used to carry gene-editing tools into cells for therapy. Higher editing power often raises the chance of unwanted DNA changes or harm to cells. Researchers focused on improving SsdAtox, a smaller enzyme that is only about two-thirds the size of common base editors. In its original form, SsdAtox works slowly and causes too much damage or toxicity.

The researchers used AlphaFold3, an artificial intelligence (AI) tool that predicts the 3D shape of proteins, to find and change a key spot (called K31) in SsdAtox. This change widened the entrance for DNA to reach the enzyme's active part, boosting its work.

A new screening system improves safety and power

They also built a testing method named Trinity-Screen. This checks three things at once in bacterial cells: good editing, few harmful DNA breaks, and low toxicity. Only enzyme versions that passed all checks advanced through many selection rounds. The best versions were then tested in human cells at many sites.

The improved SsdAtox versions showed up to 11.8 times better editing than the original, about half the unwanted breaks of earlier strong versions, ten times lower toxicity in bacteria, and up to 31 times overall gain on a new measure. One top version cut unwanted breaks by 37 percent compared to a leading editor called BE4max at several gene sites. The changes also made editing happen in a tighter, more predictable range, or editing window.

To compare tools fairly, researchers created the Base Editor Performance Index (BEPI), a score that weighs editing success against side effects. Because the new versions stay small, they could fit better into AAVs, opening doors for treating more genetic diseases.

About the Writer

More from Mindplex

Keep reading

Three more ideas worth your time.

Browse News

Discussion

Join the discussion

Sign in to share a response with the community.

Type @ to mention someone Type / or use + to add a block Highlight text, then choose Link
Loading editor

Comments cannot be edited after posting because they become part of the reputation record. Give yours a quick review first.