Can graph neural networks understand chemistry? - Dominique Beaini

If you enjoyed this talk, consider joining the Molecular Modeling and Drug Discovery (M2D2) talks live: https://valence-discovery.github.io/M...

Also consider joining the M2D2 Slack: https://join.slack.com/t/m2d2group/sh...

Title: Can graph neural networks understand chemistry?

Abstract: In this talk, I will explain how to empower graph neural networks (GNNs) for molecular property prediction with more expressive models and large datasets. Graph neural networks (GNNs) have emerged as one of the most important innovations for machine learning in drug discovery. Their ability to work on unstructured data enables us to use deep learning on molecular graphs, with the promise of predicting molecular properties with the same speed and accuracy that convolutional networks process images. However, GNNs face unprecedented challenges that I address in the talk, including the difficulty of detecting sub-structures, modeling non-covalent interactions, and the lack of large datasets.

Speaker: Dominique Beaini -   / dom_beaini  


Twitter Prudencio:   / tossouprudencio  
Twitter Therence:   / therence_mtl  
Twitter Cas:   / cas_wognum  
Twitter Valence Discovery:   / valence_ai  

~

Chapters:
00:00 Introduction speaker
00:56 In short: Deep learning for drug discovery
08:00 Molecular graphs are like a maze
16:09 Principle Neighbourhood Aggregation (PNA): More powerful aggregators
24:34 Directional Graph Networks (DGN): Positional and structural encoding
37:40 Spectral Attention Network (SAN): Graph transformers
45:30 Going beyond simple fingerprints: Transfer knowledge across tasks
53:52 Conclusion
55:25 Q&A