Predicting base editing outcomes using position-specific sequence determinants.

Area of Science:

• Genetics and Genomics
• Molecular Biology
• Bioinformatics

Background:

• CRISPR/Cas base editors offer precise DNA sequence modification.
• Understanding base editor activity determinants is crucial for effective gene editing.

Purpose of the Study:

• To investigate sequence biases influencing CRISPR/Cas base editor activity.
• To develop a predictive model for base editing efficiency and specificity.
• To assess the impact of sequence context on therapeutic gene editing outcomes.

Main Methods:

• Systematic measurement of base editing frequencies across thousands of target sequences in human cell lines.
• Training a machine learning model using sequence features to predict editing activity.
• Evaluating model performance and generalization capabilities compared to existing tools.

Main Results:

• Base editing activity is significantly influenced by flanking nucleotides, affecting the editing window.
• Position-specific sequence biases impact editing rates differently across the editing window.
• The developed machine learning model achieved prediction accuracies of 0.49–0.72 and showed improved generalization.
• Analysis of disease mutation correction guides revealed substantial unwanted editing events.

Conclusions:

• CRISPR base editing exhibits predictable, position-specific sequence biases.
• A novel machine learning model enhances the prediction of base editor activity.
• This work facilitates more efficient and precise planning of gene editing strategies for research and therapy.