CROP: a feature-independent context-aware method for CRISPR-Cas9 frameshift prediction

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• CRISPR-Cas9 technology enables precise genome editing via guide RNA-directed DNA cleavage.
• Predicting CRISPR-Cas9 repair outcomes, especially frameshifts, is crucial for effective gene editing strategies like gene knockout.
• Existing prediction methods face limitations due to sparse data, reliance on manual features, and limited experimental contexts.

Purpose of the Study:

• To develop an advanced computational method for accurate prediction of CRISPR-Cas9 repair outcomes.
• To overcome limitations of previous methods in handling sparse data and manual feature engineering.
• To improve the design of guide RNAs for predictable gene editing results.

Main Methods:

• Developed CROP, a feature-independent, context-aware repair-outcome prediction method.
• Aggregated repair outcomes into Δlength classes to address class sparsity.
• Designed CROP for variable input sequence lengths and simultaneous utilization of multiple datasets.

Main Results:

• CROP demonstrated superior performance in frameshift-rate prediction across 18 curated datasets compared to state-of-the-art methods.
• Cross-experiment and cross-cellular predictions confirmed the generalizability of CROP's learned repair mechanisms.
• CROP successfully learned microhomology principles directly from raw sequences without explicit feature engineering.

Conclusions:

• CROP provides an end-to-end, data-driven architecture for CRISPR-Cas9 repair-outcome prediction.
• The method enhances the accuracy and reliability of predicting gene editing outcomes.
• CROP facilitates more precise guide RNA design for desired genomic modifications.