Precise DNA base editing using AlphaFold3-based contact modelling

Area of Science:

• Biochemistry
• Genomics
• Artificial Intelligence

Background:

• High specificity is crucial for genome editing tools in research and therapeutics.
• Current methods face challenges with activity-specificity trade-offs, labor intensity, and low success rates.

Purpose of the Study:

• To introduce ContactSeek, an AI-driven framework utilizing AlphaFold3 (AF3) contact probability to enhance genome editor specificity.
• To demonstrate ContactSeek's ability to identify specificity-determining residues in base editors.

Main Methods:

• Mapped genome-wide off-targets of Cas9-TadA adenine base editors and input sequences into AF3.
• Utilized AF3 contact probability to detect differential interactions and correlated it with sequencing data.
• Identified consensus contact regions and specificity-determining residues.

Main Results:

• Contact probability from AF3 was more sensitive than 3D structures for detecting off-target interactions.
• ContactSeek identified key residues in Cas9 and TadA8e, leading to a variant with greatly enhanced specificity.
• The best variant outperformed several known high-fidelity adenine base editors and was generalized to Cas12a editors.

Conclusions:

• ContactSeek provides a novel AF3-driven model for improving genome editing tool specificity.
• The framework integrates structural and functional dimensions to enhance precision.
• This represents a paradigm shift in developing more accurate genome editing technologies.