CRISPR-M: Predicting sgRNA off-target effect using a multi-view deep learning network.

Area of Science:

• Molecular Biology
• Bioinformatics
• Genomics

Background:

• CRISPR-Cas9 is a powerful genome editing tool with applications in gene therapy and agriculture.
• Off-target mutations, caused by guide RNA misdirection, are a significant challenge limiting CRISPR-Cas9's precision.
• Existing computational methods for predicting off-target effects require improvement.

Purpose of the Study:

• To develop an effective computational approach for predicting CRISPR-Cas9 single-guide RNA (sgRNA) off-target effects.
• To address limitations in current off-target prediction capabilities, particularly for target sites with indels and mismatches.

Main Methods:

• Introduction of CRISPR-M, a novel approach utilizing a new encoding scheme and a multi-view deep learning model.
• Implementation of a three-branch network combining convolutional neural networks (CNNs) and bidirectional long short-term memory (BiLSTM) recurrent neural networks.
• Training and evaluation on real-world datasets to assess prediction performance.

Main Results:

• CRISPR-M demonstrated significant performance advantages over existing methods on real-world datasets.
• Experimental analysis validated CRISPR-M's capability in feature extraction for off-target effect prediction.
• The model showed superiority in predicting sgRNA off-target effects, including those with indels and mismatches.

Conclusions:

• CRISPR-M offers a superior computational solution for predicting sgRNA off-target effects in CRISPR-Cas9 genome editing.
• The developed multi-view deep learning model effectively captures complex features for accurate off-target prediction.
• This advancement holds potential for enhancing the safety and efficacy of CRISPR-Cas9 applications in various fields.