Single-particle cryo-electron microscopy (EM) is a popular technique for determining the structure of protein complexes of biomedical importance at near-atomic resolution. Unlike the problem of diffractive imaging and phase retrieval, raw cryo-EM images do contain phase information which facilitates 3D reconstruction, but since the samples are not arranged in a crystalline formation, they do not produce a diffraction pattern.
Instead, the signal from hundreds of thousands of particles has to be aligned and averaged in three-dimensions to obtain a high-resolution cryo-EM density map. This talk will explore links between coherent diffractive imaging and single-particle cryo-electron microscopy. In particular, we will comment on theoretical and practical aspects, including how to deal with big data and how to incorporate machine learning into the 3D refinement and reconstruction process.
The advent of high-throughput methods for data acquisition that can produce thousands of high-quality tilt series during a single microscope session, however, has uncovered bottlenecks in the downstream data analysis which has so far relied on supervised, user-driven pipelines.
In this talk, I will present recent advances in high-throughput tomography that have allowed us to streamline the cryo-ET structure-determination process and improve the resolution of structures using the "constrained single-particle tomography" paradigm. The combination of strategies for accelerated tilt-series acquisition together with data-driven techniques for high-resolution image analysis, will pave the way for cryo-ET to become the technique of choice to determine protein structures in their native environment.
Copyright. All Rights Reserved.