Thread by @keennethy

Very glad to share our latest review on #geometricdeeplearning in molecular sciences, with special emphasis on drug discovery, quantum chemistry and synthesis prediction. 1/6

https://twitter.com/fra_grisoni/status/1420284463469277196

We put strong focus on the various existing molecular representations and their individual symmetries and advantages for different modeling applications. 2/6

We discuss the major network architectures, such as RNNs, GNNs, 3D CNNs and Transformers, their applications on molecular systems, and the relevant equivariances they incorporate. 3/6

We place additional emphasis is on the relevance of the learned molecular features and their complementarity to well-established molecular descriptors. 3/6

Such features can be used to visualize structure-activity landscapes in drug discovery, incorporating information from the molecular structure and their biological activity. 4/6

More in depth theory about geometric deep learning can be found in the work of @mmbronstein et al. 5/6

Paper: arxiv.org/abs/2104.13478
Webpage: geometricdeeplearning.com
ICLR 2021 talk: iclr.cc/virtual/2021/i…

We delineate future opportunities in the fields of #DrugDiscovery #physchem and #structuralbiology 6/6