Predicting CRISPR-Cas9 Off-Target Effects Using Deep Learning
Author Information
Author(s): Sari Orhan, Liu Ziying, Pan Youlian, Shao Xiaojian
Primary Institution: National Research Council Canada
Hypothesis
Can a deep learning model accurately predict off-target effects of CRISPR-Cas9 in human primary cells?
Conclusion
The CrisprBERT model outperformed existing models in predicting off-target effects of CRISPR-Cas9.
Supporting Evidence
- The CrisprBERT model achieved better performance than state-of-the-art models in cross-validation tests.
- Independent testing confirmed the model's generalization ability on unseen data.
- The study expanded the training dataset to include 140 sgRNAs, improving prediction accuracy.
Takeaway
Scientists created a computer program that helps predict where CRISPR might accidentally cut DNA, which is important for safe gene editing.
Methodology
The study used a deep learning model called CrisprBERT, which combines BERT embedding and BiLSTM to predict off-target effects from sgRNA and DNA sequences.
Potential Biases
Potential biases may arise from the datasets used, which were generated under different experimental conditions.
Limitations
The model's performance may be limited by the quality and size of the training datasets.
Participant Demographics
The study focused on human primary cells, specifically T-cells.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website