/Synergy_GNN

Discovering eco-friendly hydrate inhibitors using molecular dynamics and deep learning approaches

Primary LanguageJupyter Notebook

Discovering eco-friendly hydrate inhibitors using molecular dynamics and deep learning approaches

This project was carried out at AICP2022(Artificial Intelligence Challengers Program).

📂 More information can be found here.

Abstract

  • This study shows that the combination of molecular dynamics (MD) and deep learning (DL) can be used to search eco-friendly hydrate inhibitors.
  • The MD simulation results provided detailed information about the molecules and the hydrate structures, and numerous maximum pulling force data.
  • The graph neural network (GNN) approaches were used to identify the most effective inhibitors according to their molecular graph data.
  • We believe that the combination of these methods can help to find the best possible hydrate inhibitors and predict the pulling force faster and more accurately. mm

Code information

  • DataPreprocessing.ipynb : Code for processing raw data obtained through MD simulation into graph structure data is described
  • Code.ipynb : A graph neural network model capable of learning graph structures and code for learning methods are described. It also contains a code that can visualize the binding importance of each molecule.

Research results

result