Geometric deep learning on molecular representations

Geometric deep learning (GDL) is based on neural network architectures that incorporate and process symmetry information. GDL bears promise for molecular modelling applications rely representations with different properties levels of abstraction. This Review provides a structured harmonized overview GDL, highlighting its in drug discovery, chemical synthesis prediction quantum chemistry. It contains an introduction to the principles as well relevant representations, such graphs, grids, surfaces strings, their respective properties. The current challenges sciences are discussed, forecast future opportunities attempted. becoming more important spatial structure molecules information about Kenneth Atz colleagues review progress this emerging field geometric learning.