Project Summary This proposal aims to develop computational tools that analyze the structural variability of the macromolecules imaged by Cryogenic electron microscopy (CryoEM) and Cryogenic electron tomography (CryoET). As the function of most macromolecules involves dynamic interactions among their own components or with other molecules, the structural flexibility of those macromolecules is often key to accomplishing their functions. CryoEM/CryoET makes snapshots of macromolecules embedded in vitrified ice, which provides direct information of individual protein particles in different compositional and conformational states. Using advanced computational methods, we will be able to resolve the structural heterogeneity of proteins and gain a deeper understanding of their structure-function relationship. The algorithm developed in this proposal will be using the Gaussian mixture model for protein structure representation and deep neural network for embedding snapshot images of proteins onto a latent space depicting their conformational states. In this proposal, we address the issue of protein structural variability from three aspects. First, we will build a pipeline for simultaneous orientation and conformation refinement for single particle analysis, which will make it possible to solve systems with large-scale structural variability. Second, we will integrate constraints from molecular models into our pipeline, so that prior knowledge from biochemistry can be used to guide the protein heterogeneity analysis. Finally, we will focus on CryoET and expand the method to look into the dynamic of macromolecular systems inside cells. In sum, the proposed work will produce software tools for a comprehensive analysis of protein structural variability, which will provide new insights into the functioning mechanism of macromolecules.