Title:  Cryo-electron Microscopy, Tomography, and Deep Learning on 3-dimensional Molecular Images
Abstract:  Electron cryo-microscopy (cryo-EM) and cryo-electron tomography (cryo-ET) techniques have improved dramatically over the last fifteen years. They have become powerful techniques to derive 3-dimensional molecular images of a wide range of resolutions. Many molecular complexes have been resolved to near-atomic resolutions to allow determination of atomic positions. Deep learning methods, such as AlphaFold 2, have been widely explored as an alternative method to obtain high-accuracy structures. I will discuss our recent investigations in improving a deep learning method, DeepSSETracer, through the loss function, survey of AlphaFold DB, and filament tracing in cryo-ET images.