Files in this item



application/pdf3070004.pdf (4MB)Restricted to U of Illinois
(no description provided)PDF


Title:The Energy Landscape of Associative Memory Energy Functions and Their Application to Ab Initio Protein Structure Prediction
Author(s):Hardin, Charles Corey
Doctoral Committee Chair(s):Wolynes, Peter G.
Department / Program:Biophysics and Computational Biology
Discipline:Biophysics and Computational Biology
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):Biophysics, General
Abstract:The ability to predict the structure of a protein from its amino acid sequence is one of the major practical benefits that it is hoped will result from a greater understanding of the physics of protein folding. For sequences for which a highly homologous sequence with known structure is available, this hope has largely been realized. For the so-called ab initio case, where such evolutionary information is lacking, prediction becomes less reliable. Analytical theories of the protein folding reaction reveal that the ability to fold is largely dependent on a few statistical properties of the energy landscape. These include the gap in energy between the native basin and the ensemble of misfolded states and the variance in energy of the unfolded states. This insight may be used to design model energy functions for ab initio prediction. In particular the class of associative memory models are explicitly based on energy landscape ideas. In an associative memory model the energy of a particular conformation is a function of a set of sequence-structure correlations learned from a database of known structures. By varying the content of the database it is possible to tune the energy landscape between extremes consisting of a smooth funnel to the native state on the one hand, and a rough energy surface pock-marked by local traps on the other. In addition, associative memory models based on databases which contain no globally related structures may be used to obtain ab initio predictions. The quality of predicted structures built up from purely local structural relationships in this way is sufficiently high, and the computational cost of obtaining them sufficiently low, that the technique is expected to be useful in such large scale applications as genome annotation.
Issue Date:2002
Description:91 p.
Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2002.
Other Identifier(s):(MiAaPQ)AAI3070004
Date Available in IDEALS:2015-09-25
Date Deposited:2002

This item appears in the following Collection(s)

Item Statistics