Application of the Statistical Energy Landscape Theory to Protein Structure Prediction

There has been an exponential increase in sequence information over the past few years due mostly to the large scale genome projects. The corresponding structural information, however, is being determined at a much slower rate. Consequently, the sequence to structure gap is still growing. Many scientists are looking to structure prediction through computational methods to narrow this gap. Using ideas from the energy landscape theory first introduced by Bryngelson and Wolynes, and database analysis, we have designed statistical mechanical energy functions for protein structure prediction by threading and simulated annealing molecular dynamics methods. These optimized energy functions improve upon the previous approximations of Goldstein, Luthey-Schulten and Wolynes by taking into account correlations in the energy landscape. For the self-consistent energy function used in threading, alignments are comparable to those obtain by evolutionary distance-based alignments and consistently appear to be more accurate for sequences that have lower than 20% sequence identity. The self-consistent threading algorithm we developed was used in our participation in the second annual critical assessment of protein structure prediction (CASP2) competition, and our results place us among the top three groups in comparative modeling. Using this energy function, we predicted the structures of four archaeal adenylate kinases that had less than 17% sequence similarity to any known structure. These predicted structures were then used to study the sources of protein thermostablitiy. The energy function used in the molecular dynamics approach or treatment of protein folding was optimized based on a more complete characterization of a protein's energy landscape. This characterization involved a better statistical sampling of the misfolded states and a partial treatment of dominant short-range correlations and interactions determining collapse. This lead to more accurate structure predictions than was possible by the previous treatment.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Chemistry

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2015

Author dc:creator

• Koretke, Kristin Kay

Contributors dc:contributor

• Wolynes, Peter G.

• Biology, Biostatistics

Language dc:language

eng

Identifier

(MiAaPQ)AAI9912295

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/84426