Week 09 — Applications

Three case studies: AlphaFold, drug discovery, autonomous-driving perception.

Lecture

AlphaFold 2 / 3 architecture overview · molecular property prediction at scale (Open Catalyst Project, MACE) · point-cloud perception for autonomous driving · molecular generative models.

Read before the lecture

• Jumper et al., *Highly accurate protein structure prediction with AlphaFold* (Nature 2021)

Recitation — paper discussion

Batatia et al., *MACE: Higher Order Equivariant Message Passing Neural Networks for Fast and Accurate Force Fields* (NeurIPS 2022) (paper)

Come ready to argue one side of each:

• Where does the speed advantage of MACE over DFT come from?
• What's the right benchmark for force-field accuracy in 2026?

Reference text for this week: chapter 09 of the bilingual notes — EN PDF · FR PDF.