Abstract
A deep learning algorithm maps out the continuous conformational changes of flexible protein molecules from single-particle cryo-electron microscopy images, allowing the visualization of the conformational landscape of a protein with improved resolution of its moving parts.