Use this URL to cite or link to this record in EThOS:
Title: De novo prediction of coiled-coil regions and oligomeric states
Author: Vincent, Thomas L.
ISNI:       0000 0004 2720 250X
Awarding Body: University of Bristol
Current Institution: University of Bristol
Date of Award: 2011
Availability of Full Text:
Access from EThOS:
Complexity science studies the dynamic interactions between components of a system. The sensitivity of these interactions to environmental conditions means that a single complex system expresses varying emergent behaviour; making its understanding difficult, particularly when it comes to predicting its behaviour. One such "complex" problem is protein folding, which addresses the question of how a linear sequence of amino acids carries information for a three-dimensional protein structure. Deciphering of a code, or even a subset of rules, that could accu- rately help predict the 3-dimensional arrangement of amino acids within a protein would be of tremendous benefit to the basic and applied chemical, biological and biomedical sciences. One approach to solving the protein folding problem is to learn to predict smaller structures and gradually increase the size of our targets. This Thesis follows a similar mindset and focusses on the coiled coil, a ubiquitous IX-helical protein- folding motif that is involved in many biological processes. Coiled-coil prediction can be separated into two distinct problems; firstly, the identification of coiled- coil regions within protein sequences; and, secondly, predicting the architecture and topology that a coiled-coil sequence is likely to adopt; i.e., its oligomer state and the orientation of its helices. The work presented in this Thesis is grounded on the application of statistical modelling to investigate both aspects of the coiled coil-prediction problem. First, a comprehensive assessment of the existing coiled-coil domain prediction algorithms is carried out and their limitations discussed. This forms the basis for the development of a coiled-coil domain meta-predictor, which shows improve- ments over the existing individual algorithms. The contributions of the work pre- sented in this Thesis regarding the prediction of coiled-coil topology are two-fold: firstly, SCORER 2.0, an algorithm that improves our ability to discriminate be- tween coiled-coil topologies, is introduced. Despite state-of-the-art performance, SCORER 2.0 and other existing predictors are limited to two-state predictions and cover only rv 33% of the known coiled-coil population. This triggered the devel- opment of LOGIC OIL, which is the major contribution of this Thesis. LOGICOIL is a Bayesian-based methodology that increases our ab initio prediction coverage of the known coiled-coil population from rv 33% to rv 93%, but also extends our ability to differentiate between multiple coiled-coil oligomeric states from amino- acid sequence alone. Finally, the usefulness of the prediction tools developed in this work is illustrated on problems of biological interest, more specifically on the prediction of coiled coils in the tenascin, thrombospondin and matrilin protein assemblies; i.e., proteins of the extracellular matrix. Here, the coiled-coil domain meta-predictor, and LOGICOIL are used to analyze the presence and structure of coiled-coils domains, as well as their contribution to the overall architecture of the protein complexes.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available