Deep Learning Models used for off-target predicitons in CRISPR-Cas9 gene editing

This repository includes a deep convolutional neural network for predicting the off-targets in CRISPR-Cas9 gene editing. The CNN_std conducted by TensorFlow were trained using CRISPOR dataset. The CNN_std conducted by Keras were trained on the largest sgRNA off-target dataset up to date from [1].

PUBLICATION

Please cite this paper if using our preditive model:

Jiecong, Lin. & Ka-Chun, Wong. (2018). Off-target predictions in CRISPR-Cas9 gene editing using deep learning (ECCB 2018 Proceeding Special Issue). Bioinformatics, 34(17), i656–i663. http://doi.org/10.1093/bioinformatics/bty554

PREREQUISITE

The models for off-target predicitons were conducted by using Python 2.7.13 and TensorFLow v1.4.1. Following Python packages should be installed:

• scipy

• numpy

• pandas

• scikit-learn

• TensorFlow

The Keras version of CNN were conducted by Python 3.6, TensorFlow 1.9.0, Keras 2.2.0. Following Python packages should be installed:

• scipy

• numpy

• pandas

• Keras

• TensorFlow

REFERENCE

[1] Hui Peng, Yi Zheng, Zhixun Zhao, Tao Liu, Jinyan Li; Recognition of CRISPR/Cas9 off-target sites through ensemble learning of uneven mismatch distributions, Bioinformatics, Volume 34, Issue 17, 1 September 2018, Pages i757–i765, https://doi.org/10.1093/bioinformatics/bty558

CONTAINS:

• CFDScoring/cfd-score-calculator.py : Python script to run CFD score

• predictions/cnn_std_prediction_TF.py : CNN_std conducted by TensorFlow

• predictions/cnn_std_keras.py : CNN_std conducted by Keras used for off-target prediction

---

Jiecong Lin

[email protected]

January 27 2018