Skip to content

Improving protein function prediction with synthetic feature samples created by generative adversarial networks

License

Notifications You must be signed in to change notification settings

psipred/FFPredGAN

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

95 Commits
 
 
 
 
 
 
 
 

Repository files navigation

FFPred-GAN

DOI

This is a Python implementation of FFPred-GAN method reported in

Wan, C. & Jones, D.T. (2020) Protein function prediction is improved by creating synthetic feature samples with generative adversarial networks. Nature Machine Intelligence. DOI: 10.1038/s42256-020-0222-1.


Requirements

  • Python 3.6
  • Numpy
  • PyTorch (CPU mode)
  • Scikit-learn
  • Standard computer cluster

Running

It is recommended to use a standard computer cluster that allows to run multiple jobs simultaneously (e.g. the SGE array job).

  • Step 1. Downloading real training feature samples via

    http://bioinfadmin.cs.ucl.ac.uk/downloads/FFPredGAN/RealTrainingData/.

  • Step 2. Generating positive synthetic feature samples by using the template

    ./src/Generating_Synthetic_Positive_Samples_FFPred-GAN.py.

    This template is used to train one GAN for generating synthetic protein feature samples for single GO term. In order to generate synthetic samples for mutliple GO terms, this script should be duplicated with only changing the GO term ID, or changing the absolute pathname of the training feature samples for corresponding GO terms. The GO term list can be downloaded via

    http://bioinfadmin.cs.ucl.ac.uk/downloads/FFPredGAN/GOTerm_List.txt.

  • Step 3. Running Classifier Two-Sample Tests to select the optimal synthetic feature samples by using

    ./src/Classifier_Two_Sample_Tests.py.

    This template is used to select the optimal synthetic protein feature samples generated for single GO term. In order to select the optimal synthetic samples for mutliple GO terms, this script should be duplicated with only changing the GO term ID, or changing the absolute pathname of the real and synthetic feature samples for corresponding GO terms. The selected optimal synthetic feature samples can be directly downloaded via

    http://bioinfadmin.cs.ucl.ac.uk/downloads/FFPredGAN/SyntheticTrainingData/


  • Step 4. Downloading testing feature samples and class labels via

    http://bioinfadmin.cs.ucl.ac.uk/downloads/FFPredGAN/TestingData/.

  • Step 5. Training and testing support vector machine classifier with positive synthetic feature samples augmented training data by using

    ./src/Testing_Synthetic_Positive_Samples_Augmented_SVM.py.

About

Improving protein function prediction with synthetic feature samples created by generative adversarial networks

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 4

  •  
  •  
  •  
  •  

Languages