Pre-training Molecular Graph Representation with 3D Geometry

Explore a seminar on pre-training molecular graph representation with 3D geometry presented by Hanchen Wang from the University of Cambridge. Delve into the fundamental problem of molecular graph representation learning in modern drug and material discovery. Discover how 3D geometric information plays a crucial role in predicting molecular functionalities, surpassing traditional 2D topological structures. Learn about the Graph Multi-View Pre-training (GraphMVP) framework, which leverages self-supervised learning to bridge the gap between 2D topological structures and 3D geometric views. Gain theoretical insights into the effectiveness of GraphMVP and examine comprehensive experimental results demonstrating its superior performance compared to existing graph SSL methods. This 25-minute Lennard-Jones Centre discussion group seminar, held on June 27, 2022, offers valuable knowledge for researchers and professionals in the field of molecular modeling and drug discovery.

Pre-training molecular graph representation with 3D geometry