| Single-particle cryo-electron microscopy (cryo-EM) has emerged as a powerful tool in structure determination of macromolecular complexes that are not suitable for crystallographic studies. Recent advances in direct electron detectors and image-processing techniques have allowed cryo-EM to reach near-atomic resolution (3--5 A) from small particles with low or even no symmetry. However, tools to determine structures from cryo-EM maps have lagged behind. In this dissertation we describe approaches that apply Rosetta high-resolution structure prediction methods to model-building from cryo-EM maps. Using knowledge-based information gathered from proteins of known atomic structure, we describe the development of novel tools for de novo model-building from near-atomic resolution cryo-EM maps, and for architecture determination of multi-component macromolecular assemblies from medium- to low-resolution (5--15 A) cryo-EM maps. We demonstrate that these new computational tools have shown usefulness in determining the structures from newly reconstructed cryo-EM maps, which would otherwise be difficult or impossible for human to do. These methods should enable rapid and reliable structure determination from cryo-EM maps. |
