An Attention Graph Neural Network for Stereo-active Molecules
Poster Authors
Event Type
TimeWednesday, September 15
DescriptionMolecules can show stereochemistry: two molecules with the same atomic connectivity may exhibit different bioactivity due to different spatial arrangements. We design a novel graph neural network architecture that utilizes a stereo-sensitive aggregation function and attention mechanism to improve the performance of molecular properties prediction by exploiting chiral information. Our model is interpretable by allowing visualization of the learned attention weights, providing better support for drug discovery.