While the problem of determining how a protein folds in three dimensions is essentially solved, accessing protein motions and interactions at physiological conditions is becoming more central than ever before. Indeed, proteins are flexible biological objects, constantly moving and changing their shape to interact with their environment and cellular partners. This inherent flexibility is highly relevant for protein functioning. Experimentally, it is very challenging to observe proteins directly in action at high resolution, and we mostly have access to isolated clusters of “snapshots" representative of a few functional states. In the first part of my talk, I will give an overview of the recent progress in protein structure prediction and deep learning in general. I will then present physics-based and machine-learning models developed by our team to predict protein flexibility and their functional motions. These models have proven to be very useful in predicting protein dynamics and interactions up to the cellular level or extrapolating proteins’ functional motions. I will demonstrate applications allowing the construction of multi-level representations of protein flexibility and integrative algorithms driven by low-resolution experimental observations, such as small-angle scattering.​

​