Disulfide bond information can help the prediction of protein secondary structure, tertiary structure and all-atom coordinates. Most of previous works focused on either state classification or connectivity prediction with some assumption that some constraints were added to make the problem solvable in reality. In this thesis, we propose a multi-phase approach to solve the problem. Our method can export the number of bonds and achieve 90.7% accuracy in the state classification. For the connectivity prediction problem, we use the number of bonds we predict as a base to decide bond pairs. For overcoming the ratio imbalance of samples, we propose a down-sampling method to reducing processing time. Finally, we perform the weighted graph matching algorithm to obtain the bonding pattern, which achieves 63.5% accuracy. We also achieve 48% accuracy for the thorough prediction. Our method is validated by the datasets derived from SWISS-PROT and PDB. The
results are better than the previous works.
Identifer | oai:union.ndltd.org:NSYSU/oai:NSYSU:etd-0725109-190346 |
Date | 25 July 2009 |
Creators | Chung, Wei-Chun |
Contributors | Kuo-Si Huang, Chang-Biau Yang, Chung-Lung Cho, Chia-Ning Yang, Shen-Chuan Tai |
Publisher | NSYSU |
Source Sets | NSYSU Electronic Thesis and Dissertation Archive |
Language | English |
Detected Language | English |
Type | text |
Format | application/pdf |
Source | http://etd.lib.nsysu.edu.tw/ETD-db/ETD-search/view_etd?URN=etd-0725109-190346 |
Rights | off_campus_withheld, Copyright information available at source archive |
Page generated in 0.0074 seconds