Skip to content

A graph neural network based on torch geometric for aqueous solubility prediction

Notifications You must be signed in to change notification settings

phelchegs/GNN-for-Aqueous-Solubility-Prediction-using-AqSolDB

Repository files navigation

GNN-for-Aqueous-Solubility-Prediction-using-AqSolDB

A graph neural network based on torch geometric for aqueous solubility prediction

  • Predict aqueous solubility of organic molecules using graph neural network, i.e., node (atom), edge (bond), and stacking layers, instead of chem descriptors
  • Descriptors have been presented in the repo aqueous solubility prediction. Would like to compare and draw conclusion
  • Challenges:
    • multiclass classificaiton
    • message passing includes not only node features but also edge's
    • use wandb for hyperparameters sweeping

Dependencies

  • PyTorch
  • PyTorch-Geometric
  • RDKit
  • DeepChem
  • Scikit-Learn
  • Seaborn
  • tqdm

Data Source
The well known AqSolDB: https://www.nature.com/articles/s41597-019-0151-1

About

A graph neural network based on torch geometric for aqueous solubility prediction

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published