Obtaining complete macromolecular conformational landscapes at the atomic scale by combining cryo-EM image analysis, molecular dynamics simulations and deep learning – CRYOFLEX

This project focuses on technological and methodological developments in cryo electron microscopy (cryo-EM) for structural biology, through the development of methods and software for the rapid determination of complete sets of all-atom conformations (all-atom conformational landscapes) of biomolecular complexes from cryo-EM images. Building upon the results of our established consortium, we will develop a unified and rapid methodology (and the corresponding software) that combines advanced approaches to image analysis, molecular dynamics simulation and artificial intelligence (deep learning), which will enable rapid and comprehensive determination of atomic conformational landscapes from cryo-EM images. The determination of the entire set of all-atom structural models, from millions of cryo-EM images of a biomolecule in interaction with other molecules (ligands, drugs), will have great impact on structure-based drug discovery. These developments will be tested and applied to two biomedically important macromolecular complexes exhibiting notable flexibility in the context of their binding factors related to protein synthesis (the human 80S ribosome) and protein degradation (the human ATPase p97). This will enable detailed structural analyses of these complexes and improve our understanding of the molecular mechanisms of their binding with factors, as well as their functions and dysfunctions. Furthermore, this work will serve as a reference to others aiming at obtaining all-atom conformational landscapes of large macromolecular complexes like the human ATPase p97 (~37 000 atoms) and the human 80S ribosome (~250 000 atoms), which is unamenable to classical cryo-EM image processing methods. The new developments (methods and software codes) will be applicable to any biomolecule that can be analyzed by cryo-EM.