Transformer-based anti-noise models for CRISPR-Cas9 off-target activities prediction

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• CRISPR-Cas9 gene editing offers immense therapeutic potential but is hindered by off-target effects.
• Existing prediction models often fail to fully utilize guide RNA (gRNA) and DNA sequence pair information.
• Noise within off-target datasets can compromise the accuracy of current prediction methods.

Purpose of the Study:

• To develop an advanced prediction model for CRISPR-Cas9 off-target activities.
• To enhance the exploitation of gRNA and DNA sequence pair information in off-target prediction.
• To mitigate the impact of noise in off-target datasets on prediction accuracy.

Main Methods:

• A novel coding scheme was designed, incorporating mismatch type, location, and gRNA-DNA sequence pair features.
• A transformer-based anti-noise model, CrisprDNT, was developed to address data noise.
• The model was evaluated on eight established off-target datasets.

Main Results:

• The CrisprDNT model, incorporating anti-noise loss functions, demonstrated superior performance compared to existing state-of-the-art methods.
• The novel coding scheme effectively leverages key sequence and mismatch features.
• The anti-noise approach significantly improved prediction accuracy on noisy datasets.

Conclusions:

• CrisprDNT offers a more accurate and robust method for predicting CRISPR-Cas9 off-target effects.
• The model's ability to handle noisy data makes it a valuable tool for gene editing applications.
• This work advances the reliability and practical application of CRISPR-Cas9 technology.