Design of highly functional genome editors by modelling CRISPR-Cas sequences

Area of Science:

• Genomics
• Biotechnology
• Artificial Intelligence

Background:

• CRISPR gene editors face limitations in non-native environments like human cells.
• Evolutionary constraints limit the optimization of naturally derived gene editing tools.

Purpose of the Study:

• To develop a novel, AI-designed gene editor for precise human genome applications.
• To overcome limitations of naturally occurring CRISPR systems through AI-driven design.

Main Methods:

• Trained large language models on over 1 million CRISPR operons from diverse genomic and metagenomic data.
• Generated novel Cas9-like effector proteins and tailored single-guide RNA sequences.
• Evaluated AI-generated editors for activity, specificity, and base editing compatibility.

Main Results:

• Developed OpenCRISPR-1, an AI-designed gene editor for precise human genome editing.
• Generated 4.8 times more protein clusters across CRISPR-Cas families than found in nature.
• Several AI-designed editors matched or exceeded SpCas9's performance, despite significant sequence divergence.

Conclusions:

• AI-driven design can bypass evolutionary constraints to create optimized gene editing tools.
• OpenCRISPR-1 offers a powerful, programmable, and versatile platform for gene editing.
• The release of OpenCRISPR-1 promotes ethical advancements in research and commercial applications.