Effective use of sequence information to predict CRISPR-Cas9 off-target

Area of Science:

• Genomics
• Biotechnology
• Bioinformatics

Background:

• CRISPR/Cas9 is a powerful gene-editing tool with widespread biomedical use.
• Off-target effects remain a significant challenge, limiting its practical application.
• Existing predictive models often fail to fully leverage sequence pair information for accuracy.

Purpose of the Study:

• To enhance the accuracy of predicting CRISPR/Cas9 off-target activities.
• To develop a novel model that effectively utilizes sequence pair information.
• To improve the reliability of gene-editing technologies.

Main Methods:

• Proposed a new coding scheme to represent sequence pair information, distinguishing functional regions and base/base pair types.
• Developed the CRISPR-IP model, integrating Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and attention layers.
• Trained and validated the model on two independent datasets.

Main Results:

• The proposed coding scheme effectively captures crucial sequence pair information.
• The CRISPR-IP model demonstrated superior performance in predicting off-target activities compared to existing methods.
• Validation on two datasets confirmed the model's effectiveness and generalizability.

Conclusions:

• The novel coding scheme and CRISPR-IP model significantly improve the prediction of CRISPR/Cas9 off-target effects.
• This advancement holds potential for safer and more precise gene-editing applications.
• The study provides valuable tools and insights for the gene-editing research community.