Skip to content

Template-free prediction of organic reaction outcomes

License

Notifications You must be signed in to change notification settings

Frank-LIU-520/rexgen_direct

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

23 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

rexgen_direct

Template-free prediction of organic reaction outcomes using graph convolutional neural networks

Described in A graph-convolutional neural network model for the prediction of chemical reactivity

Dependencies

  • Python (trained/tested using 2.7.6, visualization/deployment compatible with 3.6.1)
  • Numpy (trained/tested using 1.12.0, visualization/deployment compatible with 1.14.0)
  • Tensorflow (trained/tested using 1.3.0, visualization/deployment compatible with 1.6.0)
  • RDKit (trained/tested using 2017.09.1, visualization/deployment compatible with 2017.09.3)
  • Django (visualization compatible with 2.0.6)

note: there may be some issues with relative imports when using Python 2 now; this should be easy to resolve by removing the periods preceding package names

Instructions

Looking at predictions from the test set

cd into the website folder and start the Django app using python manage.py runserver. Go to http://localhost:8000/visualize in a browser to use the interactive visualization tool

Using the trained models

You can use the fully trained model to predict outcomes by following the example at the end of rexgen_direct/rank_diff_wln/directcandranker.py

Retraining the models

Look at the two text files in rexgen_direct/core_wln_global/notes.txt and rexgen_direct/rank_diff_wln/notes.txt for the exact commands used for training, validation, and testing. You will have to unarchive the data files after cloning this repo.

About

Template-free prediction of organic reaction outcomes

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Jupyter Notebook 95.0%
  • Python 1.8%
  • JavaScript 1.8%
  • CSS 1.4%