UCSF

Macromolecular assemblies are fundamental to most biological processes and here we attempt to improve the structural characterization of assemblies in the hopes that with new and improved models will produce functional insights on assemblies.

Modeling of macromolecular assemblies begins with an analysis of the computational and experimental data available on the entire assembly, subcomplexes, individual subunits and the interactions between subunits. Having collected the data on the assembly, the next challenge is to integrate the disparate data to produce a structural model. Hybrid approaches, which integrate multiple sources of data, provide a way to increase the coverage and accuracy of structure modeling for macromolecular complexes.

Fitting in with the theme of hybrid methods, Chapters 2 and 3 describe methods for modeling macromolecular assemblies, combining overall shape information (e.g., from cryo-electron microscopy) with interaction data (e.g., tandem affinity purification assays); and protein structures (or models) with NMR spectroscopy, respectively.

Chapter 4 proposes an assessment strategy for structure modeling methods that provides a way to measure how much improvement is left to be made, instead of the traditional approach of measuring how much improvement was already made. This assessment strategy provides more information on the specific limitations of the method and provides specific insight into how to best improve the method. The strategy also presents a more fair method of comparing competing methods that are assessed with different benchmark sets.

Chapter 5 describes the representation of subunits and assemblies by systems of points and restraints, explores the assumptions that underlie using points and restraints to model macromolecular structures, describes the properties of binary and multiple docking, and models the structure modeling framework.

The main contributions of this dissertation are two practical approaches for macromolecular assemblies; an assessment strategy that provides a more explicit description of the accuracy and limitation of assessed methods, improving the confidence with which the resulting models are used; and lastly, a deeper theoretical understanding of modeling macromolecular assemblies, including a path towards a more principled approach for integrating multiple sources of data.