A cocrystal prediction method of graph neural networks based on molecular spatial information and global attention

Abstract

This paper proposes a PFGNet model based on graph neural network, which adds molecular 3D point cloud information to represent the global aspect of molecules through the global attention mechanism.