MRCNN provides a deep-learning method for prediction of single-CpG-site methylation level with only local DNA sequences. It is implemented by deep learning library Tensorflow (version of tensorflow-gpu 1.2.0). It can be used for regression of continuous value of methylation and at present it is used for human methylation.
Installation
Download MRCNN by
https://github.com/TQBio/MRCNN.git
1)Installation has been tested in Linux with Python 2.7.
2)Since the package is written in python 2.7, python 2.7 with the pip tool must be installed first. MRCNN uses the following dependencies: numpy, scipy, pandas, scikit-learn, keras, tensorflow-gpu version>=1.2.0
You can install these packages first, by the following commands:
pip install pandas
pip install numpy
pip install scipy
pip install scikit-learn
pip install tensorflow-gpu
pip install keras
Run the different functions
Preprocess.py --- For Data pre-processing and extract DNA sequences with corresponding CpG loci.
Model.py --- For contribution of convolutional neural networks, MRCNN gets different structure with CNN.
Run.py --- For training the model and output the predicted values.
Evalution.py --- For evalution of model performances and including regression and classification performances.
Motifs.py --- For obtaining the conventional position weight matrices which can be visualized as sequence logos.
Figure.py --- For drawing the corresponding images to visualize the results.