Abstract

Abstract:

AlphaFold2 was recognized for its ability to rival the structural accuracy of experimental methods on single chain protein structures. How is it that this model is so successful, how was it developed and are there other similar problems out there that wait to be solved with machine learning? This talk attempts to identify some of the properties required for machine learning to be successful in a biological application. The second goal of the talk is to give a (selected) view of the current status of machine learning in biology and finally, to highlight the many remaining opportunities in this area.

Speaker Bio:

Simon is the founder of Latent Labs, a company developing generative foundation models for all molecules of life. Latent’s mission is to make synthetic biology programmable. The team is currently based in London and joined by other former members of the AlphaFold2 and DeepMind Science team. Simon has co-led DeepMind’s protein design team and set up DeepMind’s wet lab at the Francis Crick Institute in London. Before that, he was a member of the AlphaFold2 team, where he contributed to the core deep learning algorithm, including developing the uncertainty estimate that is now widely known as pLDDT.

Online link: https://meet.google.com/eod-stjw-fwe