zhenwoai / CNN-SVR

CNN-SVR: A deep learning approach for predicting CRISPR/Cas9 guide RNA on-target activity

CNN-SVR

Overview

CNN-SVR is a deep learning-based method for CRISPR/Cas9 guide RNA (gRNA) on-target cleavage efficacy prediction. It is composed of two major components: a merged CNN as the front-end for extracting gRNA and epigenetic features as well as an SVR as the back-end for regression and predicting gRNA cleavage efficiency.

Pre-requisite:

• Ubuntu 16.04
• Anaconda 3-5.2.0
• Python packages:
  numpy 1.16.4
  pandas 0.23.0
  scikit-learn 0.19.1
  scipy 1.1.0
• Keras 2.1.0
• Tensorflow and dependencies:
  Tensorflow 1.4.0
  CUDA 8.0 (for GPU use)
  cuDNN 6.0 (for GPU use)

Installation guide

Operation system

Ubuntu 16.04 download from https://www.ubuntu.com/download/desktop

Python and packages

Download Anaconda 3-5.2.0 tarball on https://www.anaconda.com/distribution/#download-section

Tensorflow installation:

pip install tensorflow-gpu==1.4.0 (for GPU use)
pip install tensorflow==1.4.0 (for CPU use)

CUDA toolkit 8.0 (for GPU use)

Download CUDA tarball on https://developer.nvidia.com/compute/cuda/8.0/Prod2/local_installers/cuda_8.0.61_375.26_linux-run

cuDNN 6.1.10 (for GPU use)

Download cuDNN tarball on https://developer.nvidia.com/cudnn

Content

• ./data: the training and testing examples with gRNA sequence and corresponding epigenetic features and label indicating the on-target cleavage efficacy
• ./weights/weights.h5: the well-trained weights for our model
• ./cnnsvr.py: the python code, it can be ran to reproduce our results

Usage

python cnnsvr.py

Note:

• The input training and testing files should include gRNA sequence with length of 23 bp and four "A-N" symbolic corresponding epigenetic features seuqnces with length of 23 as well as label in each gRNA sequence.
• The train.csv, test.csv can be replaced or modified to include gRNA sequence and four epigenetic features of interest

Demo instructions

Input (gRNA sequence and four epigenetic features):

• Data format:

• gRNA sequence: TGAGAAGTCTATGAGCTTCAAGG (23bp)
• ctcf: NNNNNNNNNNNNNNNNNNNNNNN (length=23)
• dnase: AAAAAAAAAAAAAAAAAAAAAAA (length=23)
• h3k4me3: NNNNNNNNNNNNNNNNNNNNNNN (length=23)
• rrbs: NNNNNNNNNNNNNNNNNNNNNNN (length=23)

Load weights (Pre-trained weight file):

./weights/weights.h5

Run script:

python ./cnnsvr.py

Output (Predicted activity score for gRNA):

0.22743436