A comprehensive benchmark for multiple highly efficient base editors with broad targeting scope.

Area of Science:

• Molecular Biology
• Genetics
• Bioinformatics

Background:

• Selecting optimal base editors (BEs) and guide RNAs (gRNAs) for specific gene editing targets is challenging due to an expanding toolkit.
• Understanding BE editing profiles and properties is crucial for efficient and accurate genetic modification.

Purpose of the Study:

• To construct and evaluate a comprehensive set of adenine and cytosine base editors.
• To build a predictive model for base editing efficiency and outcomes.
• To experimentally validate the use of base editors for disease-associated variants and explore large-scale variant modeling.

Main Methods:

• Constructed 10 high-activity adenine and cytosine base editors.
• Evaluated 34,040 base editor-gRNA-target combinations using long targets and tiling gRNAs.
• Developed a deep learning model (BEEP) for predicting editing efficiency and outcome.
• Experimentally validated base editor performance on 3,558 disease-associated single nucleotide variants (SNVs).

Main Results:

• Observed widespread non-canonical protospacer adjacent motifs (PAMs) for tested base editors.
• BEEP accurately predicted editing efficiency and outcome, enabling targeting of previously 'uneditable' sites.
• Successfully installed 3,558 disease-associated SNVs, including 20.1% lacking canonical PAMs.
• Predicted candidate base editor-gRNA-target combinations for 1,752,651 ClinVar SNVs.
• Identified cancer-associated SNVs linked to BRAF inhibitor resistance in melanoma.

Conclusions:

• The developed base editors and BEEP model offer a powerful resource for interrogating functional SNVs.
• This approach expands the scope of base editing, making previously inaccessible targets editable.
• A freely accessible webserver aids researchers in selecting optimal base editors and gRNAs for gene editing applications.