Predicting CRISPR/Cas9 Repair Outcomes by Attention-Based Deep Learning Framework

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• CRISPR/Cas9 is a programmable nuclease for genome editing.
• DNA mutations from CRISPR/Cas9 breaks influence biological effects.
• Repair outcomes are sequence-dependent but current prediction methods are limited.

Purpose of the Study:

• To develop a deep learning model for predicting CRISPR/Cas9 editing outcomes.
• To improve the accuracy and detail of mutation prediction.
• To understand the influence of sequence features on repair outcomes.

Main Methods:

• Developed Apindel, a deep learning model.
• Utilized GloVe for automatic sequence feature training.
• Incorporated Positional Encoding (PE) for location information.
• Employed BiLSTM and Attention mechanisms for prediction.

Main Results:

• Predicted 557 DNA repair labels, covering most Cas9 outcomes.
• Apindel demonstrated superior performance over existing models.
• Identified varying nucleotide influences based on position relative to cleavage sites.

Conclusions:

• Apindel offers accurate and detailed prediction of CRISPR/Cas9 editing outcomes.
• The model advances the understanding of sequence determinants in genome editing.
• This work supports the clinical application of CRISPR/Cas9 technology.