Team uses AlphaFold AI to redesign gene-editing proteins to make them safer

2026-07-24 · Source: AI - Ars Technica · Field: Science & Research — Life Sciences & Biology, Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Expert, medium

Summary

Researchers have successfully modified the AlphaFold AI protein-folding software to redesign gene-editing proteins, significantly enhancing their safety by reducing off-target effects. The team, based in China, developed a method called "ContactSeek" to identify specific amino acids in Cas9 proteins that mediate problematic interactions with unintended DNA sequences. After building a library of off-target sites and modeling Cas9 interactions using AlphaFold, they compared on- and off-target structures to pinpoint regions where amino acids altered contacts during mismatches. By focusing on clusters of these amino acids, they made 23 different swaps at 10 key positions. This led to a Cas9 variant that reduced off-target activity from 28 percent to 5 percent while maintaining on-target efficacy. This approach also proved effective for Cas12 and offers a general method for tailoring gene-editing systems to prevent specific off-target events, with broader implications for protein-DNA interaction fine-tuning.

Key takeaway

For AI Scientists and Research Scientists developing gene therapies, this work offers a powerful new strategy to enhance safety and specificity. You should consider integrating AI-driven protein modeling, specifically modified AlphaFold and "ContactSeek," into your design workflows to precisely identify and mitigate off-target editing risks. This method allows for tailoring gene-editing systems to prevent known off-target events, potentially accelerating the development of safer and more effective therapeutic applications by fine-tuning protein-DNA interactions.

Key insights

Modified AlphaFold can precisely redesign gene-editing proteins to drastically reduce off-target effects, enhancing therapeutic safety.

Principles

• Cas proteins tolerate some DNA mismatches.
• Off-target interactions alter protein contact points.
• AI modeling can predict problematic protein regions.

Method

Researchers used AlphaFold to model Cas9-DNA/RNA interactions, identified amino acids with altered contact probabilities during mismatches via "ContactSeek," then modified these regions to improve specificity.

In practice

• Use AlphaFold for protein-nucleic acid interaction analysis.
• Apply "ContactSeek" to identify off-target interaction sites.
• Tailor gene-editing systems for specific off-target prevention.

Topics

• AlphaFold
• Gene Editing
• CRISPR-Cas9
• Off-target Effects
• Protein Engineering
• Computational Biology

Best for: AI Scientist, Research Scientist

Related on AIssential

• CRISPR gets a power boost from AI-designed ‘molecular scissors’ — AI-designed synthetic CRISPR nucleases offer enhanced genome editing efficiency, overcoming natural protein design limitations.
• Our approach to bioresilience — AI models are crucial for global biosecurity, enabling proactive defense, rapid detection, and accelerated response to biological threats.
• AlphaFold: Grand Challenge to Nobel Prize with John Jumper — AlphaFold's success demonstrates AI's profound utility in accelerating scientific discovery, even without full interpretability.
• From discovery to design: in conversation with Ali Madani (Profluent) — AI-driven language models can design novel, functional proteins from scratch, accelerating biological evolution for therapeutic applications.
• The Dog That Hacked Cancer — AI tools can accelerate personalized medicine development, even for non-experts.
• Bio-Techne Launches Expanded R&D Systems AI-Engineered Designer Protein Portfolio — AI-engineered proteins enhance stability and activity, improving reproducibility and scalability in advanced cell culture and therapy.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI - Ars Technica.