Efficient C•G-to-G•C base editors developed using CRISPRi screens, target-library analysis, and machine learning

Area of Science:

• Molecular Biology
• Genetics
• Bioengineering

Background:

• Programmable cytosine-guanine to guanine-cytosine (C•G-to-G•C) base editors (CGBEs) offer significant scientific and therapeutic promise.
• However, predicting C•G-to-G•C editing outcomes and achieving high efficiency and purity remain challenges.

Purpose of the Study:

• To develop engineered CGBEs and machine learning models for efficient and high-purity C•G-to-G•C base editing.
• To identify factors influencing C•G-to-G•C editing through a CRISPR interference screen.

Main Methods:

• A CRISPR interference (CRISPRi) screen targeting DNA repair genes was conducted.
• Ten engineered CGBEs were characterized across 10,638 genomic sites in mammalian cells.
• Machine learning models were trained to predict editing purity and yield.

Main Results:

• Machine learning models accurately predicted editing outcomes (R=0.90).
• Engineered CGBEs achieved >90% precision (mean 96%) and up to 70% efficiency (mean 14%) for disease-related variants.
• Computational prediction enabled high-purity editing at over fourfold more sites compared to single CGBE variants.

Conclusions:

• The developed CGBE suite and predictive models significantly enhance C•G-to-G•C base editing efficiency and precision.
• This approach facilitates the correction of numerous disease-related genetic variants.
• Optimized CGBE-single-guide RNA pairing expands the scope of precise base editing applications.