Investigation into protein folding prediction of helices using techniques in computer science /
The research presented in this dissertation focuses on the application of computer science techniques in the field of theoretical biochemistry. This interdisciplinary study analyzes current black-box neural network systems and applies information from the analysis into a novel step-wise (white-box)...
| Main Author: | |
|---|---|
| Format: | Thesis Book |
| Language: | English |
| Published: |
[Place of publication not identified] :
[publisher not identified] ;
1998.
|
| Subjects: | |
| Online Access: | http://proxy.library.tamu.edu/login?url=http://proquest.umi.com/pqdweb?did=737708531&sid=1&Fmt=2&clientId=2945&RQT=309&VName=PQD |
| Summary: | The research presented in this dissertation focuses on the application of computer science techniques in the field of theoretical biochemistry. This interdisciplinary study analyzes current black-box neural network systems and applies information from the analysis into a novel step-wise (white-box) Bayesian prediction system with heuristic refinement that provides insight into the prediction and performs comparably with existing prediction models. This research studies the prediction process that determines a protein's helices from the primary amino acid sequence. Existing neural network prediction systems are analyzed and some of the factors that the systems consider important in the prediction process are identified. This information is then applied in a step-wise Bayesian prediction system and refined using heuristics that incorporate high-level knowledge of the helix structure. The result is a white-box prediction system that is at least as accurate as the black-box system and provides insight into the prediction process. The Bayesian prediction system used in this research focuses on the prediction of helices by using region-specific, position-dependent helical propensities. Because variations in local amino acid sequences determine whether a helix is formed, we infer that each amino acid has explicit preferences toward specific regions (N-, C- terminals, and middle) in helices. These region-specific, position-dependent propensities appear to correlate with spatial organization along the helix wheel. Furthermore, the statistical analysis indicates that the helix propensities are conditionally independent for structural determination. Using this information, a statistical approach is proposed for determining helix high-level helical patterns with knowledge-based postprocessing provides a novel step-wise approach to helix location identification. |
|---|---|
| Item Description: | Vita. "Major Subject: Computer Science". |
| Physical Description: | xii, 127 leaves : illustrations ; 28 cm. Issued also on microfiche from University Microfilms Inc. |
| Bibliography: | Includes bibliographical references: pages 95-99. |