A team in Melbourne has used artificial intelligence to design molecules that can quickly and precisely switch off CRISPR, paving the way for safer gene-editing therapies. The new method could accelerate progress in medicine, agriculture and basic research.
When a 10-month-old baby in the United States had a rare genetic disease effectively cured with CRISPR last year, it marked a turning point for gene editing. But it also underscored a major challenge: how to harness CRISPR’s power without risking damage to healthy DNA.
Now, scientists in Melbourne say artificial intelligence could help solve that problem.
A team led by Gavin Knott, an associate professor at the Monash University Biomedicine Discovery Institute, working with D. Rhys Grinter at the University of Melbourne’s Bio21 Molecular Science and Biotechnology Institute, has developed an AI-driven method to design molecules that can precisely shut down CRISPR when needed.
Their experimental study, published in Nature Chemical Biology, describes a way to create “anti-CRISPR” proteins in a matter of weeks, potentially making gene editing safer and more controllable.
CRISPR is often described as a pair of genetic scissors. It allows scientists to cut, remove and replace faulty genes with remarkable accuracy. But once activated inside a cell, CRISPR’s cutting enzyme can linger, sometimes snipping at the wrong spots in DNA or RNA. These so-called off-target effects can introduce harmful mutations in otherwise healthy genes.
To keep CRISPR in check, researchers have turned to anti-CRISPRs — natural proteins originally discovered in viruses that infect bacteria. These viral proteins evolved to block CRISPR defenses, and scientists realized they could be repurposed as off switches for gene-editing tools.
The catch is that natural anti-CRISPRs are rare and difficult to find. Over an entire decade of CRISPR research, only 118 such molecules have been identified.
“[T]he discovery of natural inhibitors against clinically relevant targets remains challenging and time-consuming,” Grinter said in a news release.
That bottleneck is where AI comes in.
Instead of searching nature for the right anti-CRISPR, the team used artificial intelligence to design new ones from scratch.
“[U]sing AI-accelerated protein design, we rapidly produced functional inhibitors of CRISPR that function in bacterial and human cells,” added lead author Cyntia Taveneau, a research fellow in the Knott lab.
In other words, the researchers trained computational models to predict protein shapes that could latch onto CRISPR enzymes and block their activity. They then built and tested these designer proteins in the lab, confirming that they could control CRISPR in both simple and more complex cellular systems.
The study focused on CRISPR–Cas13, a type of CRISPR system that targets RNA rather than DNA. RNA editors like Cas13 are being explored for treating diseases where changing RNA messages, instead of permanently altering DNA, may be safer or more flexible.
To control this RNA editor, the team used AI to generate highly specific anti-CRISPRs that act as molecular brakes.
“In this study we implemented a rapid approach to anti-CRISPR design that uses AI to create highly accurate and specific anti-CRISPRs, in this case to control the activity of an RNA editor,” Grinter added.
Crucially, the new approach is fast. The process took about eight weeks from choosing a target to identifying promising anti-CRISPR candidates — a dramatic speedup compared with traditional protein discovery, which can take many months or even years.
That acceleration could be important as CRISPR-based therapies move closer to routine clinical use. Having a reliable way to dial gene-editing activity up or down, or to switch it off entirely, could make treatments safer and easier to fine-tune for individual patients.
The capability to “design bespoke inhibitors that can keep CRISPR ‘in line’ will contribute to the ongoing development of CRISPR tools in diverse applications across research, medicine, agriculture and microbiology,” added Knott.
In medicine, AI-designed anti-CRISPRs could act as safety switches for therapies that edit genes in the body, helping to limit off-target effects and reduce long-term risks. In agriculture, they could give researchers more precise control over gene-edited crops or livestock. In basic science, they offer a powerful way to study how CRISPR systems work by turning them on and off at will.
The work also highlights a broader shift in biotechnology: instead of only discovering useful molecules in nature, scientists are increasingly designing them with the help of AI. Protein design tools can scan vast numbers of possible shapes and sequences, narrowing down candidates that are most likely to work before any lab experiments begin.
For students and early-career researchers, the study is a glimpse of where the field is heading. Future gene-editing tools may come paired with built-in, AI-designed safeguards, making powerful technologies like CRISPR more predictable and acceptable for clinical use.
As CRISPR continues to move from the lab bench to the clinic, the ability to control it with precision may be just as important as the ability to cut and edit genes in the first place. With AI-designed anti-CRISPRs, researchers are beginning to build that control into the technology from the ground up.
Source: Monash University