This project provides support for methods development in Cryo Electron Microscopy (CryoEM) and cellular
Cryo Electron Tomography (CryoET), with a focus on software methods for extracting maximal biological
information from the collected image data. Our methods development is driven by specific diverse and
important biological projects. When imaging, electron beam causes damage to the specimen as it is being
imaged, so very low doses must be used, leading to images with extremely high relative noise levels.
Additionally, the targets being imaged are typically individual molecules or macromolecular assemblies,
~10-30 nm in size, >10x smaller than the wavelength of visible light. These targets are ostensibly identical,
in that they all represent the same molecular species, but when imaging inside vitrified cells, the targeted
macromolecules will typically be interacting with other molecules and often undergoing significant dynamics.
Even when studying purified molecules in CryoEM, the molecules represent their solution configuration,
frozen in a single instant in time, and each will also be in a random 3-D orientation. Image processing must
then take these very noisy 2-D images of 3-D molecules in random orientations, with compositional and
conformational variability, and produce one or more high resolution 3-D structures. This proposal seeks to
improve the methods used to perform these tasks, apply the developed methods to important biological
projects, and disseminate the methods to CryoEM/ET users around the world to help solve additional
biological problems. A range of different methods will be developed, and we will adapt to the needs of the
field over time. Deep learning approaches combined with Gaussian modeling are providing promising
results in simultaneously characterizing molecular variability and structure, and there are hopes that this
method may eventually extend to near-atomic resolution. In cells, a key need is to annotate subcellular
features and identify specific macromolecules of interest. Valuable biological information can come from the
annotation of overall features, the structures of macromolecules within the cell, and by combining this
information; that is, not just studying all macromolecules of a particular type, but also considering where
individual macromolecules are within the cell, relative to other cellular components. The new methods we
are developing will enable cell and structural biologists to study cells in new ways, asking questions about
molecular interactions and functional conformational changes of macromolecules in a cellular context.