Session #6
- Date: Wednesday, 14th December 2022
- Time: 15h00 - 17h00
- Hybrid event (live on SCAI meeting room and also online)
- Pre-register here
- Link to the online event.
Talks:
Deep Learning for protein structure prediction: focus on Alphafold (slides)
By: Thibaut VĂ©ry - Idris/CNRS
Proteins are important macromolecules fulfilling all kinds of roles in living organisms. Their structure involve a sequence of aminoacids (small building blocks) with a precise organisation in space that wil define the role. Getting the sequence is quite “easy” but it is not sufficient. We need to know the 3D structure to be able to tell what the protein does, we talk about the folding of the sequence.
Several experimental methods are available to get the structure but they might not work for all proteins and are quite expansive.
Since 1994, the CASP contest, holds every 2 years to compare computational methods to find the 3D configuration of unknown proteins. Several machine learning methods are used in these contests and in 2018 DeepMind introduced the Deep Learning model Alphafold which was much better than all the other methods.
This talk will introduce the problematic of protein folding and discuss the architecture of Alphafold to explain why there was this huge gain in performance.
Deep Learning Optimisé sur Jean Zay (slides)
By: Bertrand Cabot - Idris/CNRS