Topological models of transmembrane proteins for subcellular localization prediction

Maetschke, Stefan (2007). Topological models of transmembrane proteins for subcellular localization prediction PhD Thesis, School of Information Technology and Electrical Engineering, University of Queensland.

       
Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials)
Name Description MIMEType Size Downloads
n01front_maetschke.pdf n01front_maetschke.pdf application/pdf 913.08KB 7
n02content_maetschke.pdf n02content_maetschke.pdf application/pdf 27.69MB 5
Author Maetschke, Stefan
Thesis Title Topological models of transmembrane proteins for subcellular localization prediction
School, Centre or Institute School of Information Technology and Electrical Engineering
Institution University of Queensland
Publication date 2007
Thesis type PhD Thesis
Abstract/Summary Transmembrane proteins are proteins that are inserted into the membranes of the cell and its organelles. They perform a variety of essential functions as channels, pumps, receptors and energy transducers, and are therefore a major target for drug development. It is estimated that about 20%-30% of a proteome are transmembrane proteins but structure and function have been identified for only a fraction of them. Proteins are distributed to their specific subcellular locations within the cell by highly complex sorting machinery. A crucial first step in revealing the function of a protein is therefore the determination of its subcellular localization. Experimental localization techniques are time consuming and expensive and cannot keep pace with the exponentially growing number of protein sequences. Consequently, computational techniques for subcellular localization prediction have been developed. The vast majority of these prediction algorithms are designed for soluble proteins however, and ignore the characteristic topology of transmembrane proteins. In this thesis, topological models of transmembrane proteins for subcellular localization prediction are constructed and studied, utilizing three different machine learning techniques: Support Vector Machines, Hidden Markov Models and Conditional Random Fields. The objectives of the thesis are: Firstly, to evaluate whether topological models achieve higher prediction accuracies than non-topological models. Secondly, to examine methods to model the topological regions of transmembrane proteins. Thirdly, to study the relationship between topological regions and protein localization. And fourthly, to compare different machine learning techniques in the context of transmembrane protein modeling and subcellular localization prediction. The prediction performances of the constructed models are evaluated on two datasets. The first one is a subset of the mouse proteome, containing transmembrane proteins with experimentally confirmed subcellular localizations. The second one is a selection of eukaryotic transmembrane proteins, extracted from the Swiss-Prot database.

 
Citation counts: Google Scholar Search Google Scholar
Access Statistics: 192 Abstract Views, 12 File Downloads  -  Detailed Statistics
Created: Fri, 21 Nov 2008, 15:36:18 EST