/mol_mpnn

Efficient Learning of Message Passing Neural Networks for Molecular Property Production

Primary LanguagePython

mol_mpnn

Tensorflow implementation for efficient learning of message passing neural networks for molecular property production

Components

  • preprocessing.py - script for preprocessing data
  • train.py - script for model training
  • test.py - script for model evaluation (molecular property prediction)
  • MPNN.py - model architecture

Dependencies

  • Python
  • TensorFlow
  • RDKit
  • NumPy
  • scikit-learn
  • sparse

Results (Mean Absolute Error on Test Set)

Property MAE
mu 0.0293
alpha 0.0131
HOMO 0.0557
LUMO 0.0271
gap 0.0395
R2 0.0049
ZPVE 0.0029
U0 0.0013
U298 0.0013
H298 0.0015
G298 0.0013
Cv 0.0117