Skip to content

A hybrid deep convolutional neural network for predicting chromatin accessibility

License

Notifications You must be signed in to change notification settings

kimmo1019/Deopen

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

49 Commits
 
 
 
 
 
 
 
 

Repository files navigation

Deopen

Deopen is a hybrid deep learning based framework to automatically learn the regulatory code of DNA sequences and predict chromatin accessibility.

Requirements

  • h5py
  • hickle
  • Scikit-learn=0.18.2
  • Theano=0.8.0
  • Lasagne=0.2.dev1
  • nolearn=0.6.0

Installation

Download Deopen by

git clone https://github.com/kimmo1019/Deopen

Installation has been tested in a Linux/MacOS platform with Python2.7.

Instructions

Preprocessing data for model training

python Gen_data.py <options> -pos <positive_bed_file> -neg <negative_bed_file> -out <outputfile>
Arguments:
  positive_bed_file: positive samples (bed format)
  e.g. chr1	9995	10995	
       chr3	564753	565753
       chr7	565935	566935
       
  negative_bed_file: negative samples (bed format)
  e.g. chr1	121471114	121472114	
       chr2	26268350	26269350
       chr5	100783702	100784702
  
  outputfile: preprocessed data for model training (hkl format)
 
Options:
  -l <int> length of sequence (default: 1000)

Run Deopen classification model

THEANO_FLAGS='device=gpu,floatX=float32' python Deopen_classification.py -in <inputfile> -out <outputfile>
 Arguments:  
  inputfile: preprocessed data for model training (hkl format)  
  outputfile: prediction outcome to be saved (hkl format)

Run Deopen regression model

THEANO_FLAGS='device=gpu,floatX=float32' python Deopen_regression.py -in <inputfile> -reads <readsfile> -out <outputfile>
 Arguments:  
  inputfile: preprocessed file containing different features (hkl format)  
  readsfile: reads count for each sample (hkl format)  
  outputfile: trained model to be saved (hkl format)

Citation

Liu, Qiao, et al. "Chromatin accessibility prediction via a hybrid deep convolutional neural network." Bioinformatics 34.5 (2017): 732-738.

@article{liu2017chromatin,
  title={Chromatin accessibility prediction via a hybrid deep convolutional neural network},
  author={Liu, Qiao and Xia, Fei and Yin, Qijin and Jiang, Rui},
  journal={Bioinformatics},
  volume={34},
  number={5},
  pages={732--738},
  year={2017},
  publisher={Oxford University Press}
}

License

This project is licensed under the MIT License - see the LICENSE.md file for details

About

A hybrid deep convolutional neural network for predicting chromatin accessibility

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages