CRISPR-DIPOFF (Deep Interpretable Predictors for OFF-targets)

Introduction

CRISPR-DIPOFF is an Interpretable Deep Learning Approach for CRISPR Cas-9 Off-Target Prediction. It has outperformed the previous studies by a significant margin. The related paper can be found here.

Released Models

1. LSTM model trained with 4-channel encoding
2. Finetuned ELECTRA model (Will be published soon)

Usage for Prediction

Requirements

• Python 3.8
• pandas 1.5.3
• numpy 1.22.4
• pytorch 2.0.1+cu118
• notebook 6.4.8

Prediction with LSTM Model

1. Put the input file in Sample Input directory
2. Follow the input format of Sample Input/sample_input.csv.
3. Run the notebook Predict with LSTM Model.ipynb.

Prediction with Finetuned ELECTRA Model

Experiments are ongoing. Will be published soon.

Experiments

RNN-Based Experiments with Genetic Algorithm

Data available in RNN Experiments/Data/Sequence Scripts available in RNN Experiments Steps:

1. Run Preprocessing Script RNN Experiments/Preprocessing.ipynb
2. Run Genetic Algorithm RNN Experiments/Genetic_Algorithm_RNNs V3.ipynb
3. Run Model Interpretation RNN Experiments/BEST_Model_Interpretation.ipynb
4. Run Interpretation Visualizer RNN Experiments/Data Analysis.ipynb

ELECTRA-Based Experiments

Experiments are ongoing. Will be published soon.

Citation

If you use this code for publication, please cite the original paper.

@article{toufikuzzaman2024crispr,
  title={CRISPR-DIPOFF: an interpretable deep learning approach for CRISPR Cas-9 off-target prediction},
  author={Toufikuzzaman, Md and Hassan Samee, Md Abul and Sohel Rahman, M},
  journal={Briefings in Bioinformatics},
  volume={25},
  number={2},
  pages={bbad530},
  year={2024},
  publisher={Oxford University Press}
}

Contact

For help or issues using this codebase, please submit a GitHub issue.

For personal communication related to this codebase, please contact Md. Toufikuzzaman (md.toufikzaman@gmail.com).