Advances in electron microscopy (EM) allow for structure determination of large macromolecular machines at increasingly high resolutions. A key step in this process is interpreting the EM density map with structural models of maximal accuracy and optimal precision. Model precision should be determined by the uncertainty in the experimental data; however, current methods only set uncertainty in an ad hoc manner with arbitrary weight terms. Thus, there is a need for more objective methods.
In Chapter 2, I present a novel Bayesian approach to modeling macromolecular structures based on EM density maps. The key advancement is the development of a scoring function that uses the local uncertainty of the density map to set the data weight and allows for correlation between neighboring density values. Unlike traditional approaches, the score does not require an expert user to set arbitrary parameters. I assessed the accuracy of models generated by this approach with a set of experimentally-derived, previously-published EM data of macromolecular complexes at varying resolutions from 3 to 6Å. I found that this approach leads to higher fluctuations in the model ensemble in locations with higher local uncertainty, and obtained accurate ensembles for a 3.2Å resolution map of Trpvl and 3.4Å and 5.4Å resolution maps of γ-secretase.
In Chapter 3, I present models of the γ-tubulin small complex in two functional states based on a challenging data set consisting of low-resolution EM density maps and a remotely related structure. Here, I used a traditional scoring techniques, but extensively sampled alignments and conformations in order to ensure that the model ensemble reflected the uncertainty in the data. The resulting models form a tight cluster for each state and were consistent with a set of newly reported chemical cross-links. Comparing the two states, I found significant structural differences and predict stabilizing interactions of the two states. The work in this chapter shows the difficulties of traditional modeling and serves as motivation for the methods developed in Chapter 2.
Both approaches are incorporated into the open source Integrative Modeling Platform (IMP) package, enabling integration with multiple other data types and usage of myriad sampling and analysis tools.
|Commitee:||Agard, David A., Cheng, Yifan|
|School:||University of California, San Francisco|
|School Location:||United States -- California|
|Source:||DAI-B 78/03(E), Dissertation Abstracts International|
|Keywords:||Bayesian inference, Computational biology, Computational structure prediction, Electron microscopy, Molecular dynamics, Structural biology|
Copyright in each Dissertation and Thesis is retained by the author. All Rights Reserved
The supplemental file or files you are about to download were provided to ProQuest by the author as part of a
dissertation or thesis. The supplemental files are provided "AS IS" without warranty. ProQuest is not responsible for the
content, format or impact on the supplemental file(s) on our system. in some cases, the file type may be unknown or
may be a .exe file. We recommend caution as you open such files.
Copyright of the original materials contained in the supplemental file is retained by the author and your access to the
supplemental files is subject to the ProQuest Terms and Conditions of use.
Depending on the size of the file(s) you are downloading, the system may take some time to download them. Please be