… 3D tasks. In this work, we tackle 3D molecular pretraining in a complete and novel sense. In
particular, we first propose to adopt an equivariant energy-based model as the backbone for …

… by 3D geometry due to its encoded energy knowledge, we aim to make use of the 3D
geometry of molecules in pre-training. … Unite: Unitary n-body tensor equivariant network with …

… stage, we can potentially encode more accurate molecular geometry information on the 2D
molecular graph, … Equivariant message passing for the prediction of tensorial properties and …

… In contrast to previous methods, this paper proposes one pretrained model for multiple
domains (see Figure 1), to enable unified geometric learning on 3D molecules. We claim several …

… most broadly applicable to molecular structure—on these … pretrained equivariant
representations can boost performance on downstream tasks. These results suggest that equivariant …

… [30] introduce equivariant networks to ensure the … only leverage the 2D molecular graphs for
pre-training, our method … This shows the effectiveness of using 3D information in pre-training…

… and Clebsch-Gordan decomposition to build equivariant functions, but we explicitly design
functions that are equivariant and operate on 3D-Cartesian coordinates for faster and more …

… (3)-EGMN [ 27 ] as the molecule encoder network for the ProtMD … features and 3D coordinates,
while the pre-training stage … the 3D coordinates is used in the self-supervised pre-training …

… To leverage this advantage better, we incorporate an equivariant graph neural network (GNN)
block into the architecture, inspired by [38, 39], to efficiently encode crucial information …

… domain of 3D atomistic graphs such as molecules even when 3D-… well to 3D atomistic graphs
and present Equiformer, a graph … ing SE(3)/E(3)-equivariant features based on irreducible …